Algorithms doi: 10.3390/a19060427

Authors: Jie Li Usman Adeel Muhammad Safwan Akram

Healthcare analytics is often limited by data silos and strict privacy requirements, which make it difficult to share patient-level records across organisations and to build robust predictive models. Federated learning (FL) provides an alternative by keeping data local and exchanging model updates instead of raw records. However, many existing FL solutions remain difficult to deploy in healthcare settings, as they provide limited support for auditability, governance-oriented evidence, and system-level transparency. This paper presents MediVault, an auditable and security-aware federated learning-based system for privacy-preserving healthcare collaboration. MediVault combines round-based federated training, prototype-level protected update exchange, audit-ready telemetry, and an interactive dashboard that exposes non-sensitive evidence of collaboration, model progress, and protocol execution. In addition, the system supports controlled reporting to improve stakeholder communication during pilot deployments. We evaluate MediVault on two public healthcare classification datasets, Breast Cancer Wisconsin (Diagnostic) and Heart Disease, under IID and label-skewed Non-IID settings. Experiments are conducted using logistic regression, linear SVM, and an additional lightweight MLP under matched settings. The observed results suggest that federated training remains competitive with centralised training under the evaluated settings. A prototype-level overhead analysis further shows that protected update exchange introduces measurable computational and communication costs, especially for larger update vectors. These findings indicate that MediVault can support initial system-level validation of auditable, privacy-preserving healthcare FL workflows, while further work is needed for larger-scale deployment, stronger adversarial evaluation, and real-world clinical validation.

Algorithms doi: 10.3390/a19060426

Authors: Xianbang Sun Xue Liu Yumeng Cao Chunyu Liu

Several gene-based tests, such as the sequence kernel association test, have been developed to assess associations between rare single nucleotide variants (SNVs) and disease traits. However, these aggregate methods do not distinguish potentially causal variants from null variants within associated regions. To address this limitation, we propose gvClust, a clustering approach that classifies rare variants into null and signal groups using a Gaussian mixture model applied to variant-level summary statistics from multiple-variant models. Signal variants are further partitioned into risk and protective subgroups according to their effect direction and magnitude. We evaluated gvClust in simulation studies using the adjusted Rand index (ARI), mean squared error (MSE), and accuracy of cluster number selection under different sample sizes, effect configurations, outcome types, and linkage disequilibrium (LD) structures. In simulations, gvClust showed improved performance with increasing sample size, achieved high accuracy in determining the number of clusters for continuous traits at large sample sizes, and outperformed both k-means clustering and initialization-only clustering. We then applied gvClust to rare variants in six genes associated with blood pressure traits from a large genome-wide association study and meta-analysis. In the real-data application, gvClust identified distinct null, risk, and protective clusters. These results suggest that gvClust provides a practical framework for classifying rare variants within associated regions and may help improve the biological interpretation of rare variant signals.

Algorithms doi: 10.3390/a19060425

Authors: Tingxi Wen Hanxiao Jiang Xinwen Chen Yuqing Fu Minyu Zheng

To achieve an optimal trade-off between production efficiency and energy benefits in complex manufacturing environments, this paper addresses the Green Flexible Job Shop Scheduling Problem (GFJSP) by establishing a multi-objective mathematical model that minimizes both makespan and total energy consumption. An Improved Non-dominated Sorting Genetic Algorithm II (INSGA-II) is proposed to solve this model. In the population initialization phase, chaotic mapping is integrated with multiple heuristic rules to generate a high-quality and uniformly distributed initial population. Furthermore, an enhanced elite selection mechanism is employed to effectively prevent premature convergence. Subsequently, adaptive crossover and mutation operators are designed to enable differentiated evolution across sub-populations, effectively coordinating global exploration and local exploitation. Finally, experimental results on the Brandimarte and Hurink benchmark datasets demonstrate the superiority of the proposed algorithm in terms of convergence and diversity, providing a robust solution for optimizing green industrial production scheduling.

Algorithms doi: 10.3390/a19060424

Authors: Cui Yu Liu Li Huang Sun Wang

As the Industrial Internet of Things becomes increasingly interconnected with critical infrastructure, intrusion traffic exhibits characteristics such as high-dimensional redundancy, class imbalance, and temporal correlation, posing challenges for detection systems in terms of feature representation, model complexity control, and real-time performance. To address the aforementioned issues, this paper proposes an intrusion detection algorithm based on feature optimization and asynchronous advantage actor-critic learning (APA3CID). First, the raw dataset was preprocessed using methods such as label encoding and normalization. Feature selection was performed using the improved Whale Optimization Algorithm (WOA) to reduce data redundancy and eliminate irrelevant features. The samples were then serialized based on the order in which they were collected. Second, we model the detection process as a Markov decision process, use a sliding window to construct states that capture recent temporal features, and, building upon the Asynchronous Advantage Actor-Critic (A3C) framework, we incorporate an adaptive exploration mechanism to address the issues of insufficient exploration in the early training phase and unstable convergence in the later phase. Additionally, we introduce an asynchronous lag correction strategy that utilizes truncated importance weights to mitigate the bias caused by policy lag in asynchronous parallel training, thereby enhancing the stability and robustness of policy updates. Finally, experimental results show that on the X-IIoTID dataset, APA3CID achieves a 3.51% increase in detection rate and a 4.26% increase in F1-score compared to the traditional A3C algorithm. On the WUSTL-IIoT-2021 dataset, single-sample prediction takes as little as 11.56 microseconds, with Acc, DR, and F1-score all exceeding 90%. This outperforms comparison models such as LR, XGBoost, CNN, and the baseline A3C, meeting the requirements of industrial IoT scenarios for low false-negative rates and high real-time performance.

Algorithms doi: 10.3390/a19060423

Authors: Kostiantyn Hrishchenko Oleksii Pysarchuk

This paper addresses the flexible job-shop scheduling problem (FJSP) as a constrained combinatorial optimization task with a large discrete action space. Although action-masked reinforcement learning has shown promise for such problems, the effect of structured vector-state encoding in scheduling has received less attention. The main contribution of this work is a structured block-wise state representation and a multi-branch feature extraction module for action-masked Proximal Policy Optimization (PPO). The proposed representation decomposes the scheduling state into three heterogeneous components capturing resource availability, operation readiness, and temporal attributes of operation–machine alternatives. Instead of flattening these signals into a single vector, the proposed encoder processes each block separately before aggregation, with the aim of preserving semantic structure during policy learning. To isolate the effect of representation design, we compare the proposed multi-branch encoder with a baseline single-branch multilayer perceptron under identical PPO hyperparameters and training conditions. Experiments on the Brandimarte MK benchmark suite show that the proposed architecture yields a lower best-achieved makespan on nine of ten instances and improves the best baseline result by up to 27.84%. Additional validation on selected Behnke and Geiger instances indicates that the BR encoder’s advantage extends to larger FJSP cases while preserving sub-second inference.

Algorithms doi: 10.3390/a19050422

Authors: Jichen Yuan Zepeng Su Zhulin Liu

Aiming at the challenges of class imbalance, tiny defect scales, and complex brushed background interference in the surface defect detection of electrical commutators, this paper proposes a high-precision and lightweight improved instance segmentation algorithm named WG-YOLOv11. Firstly, to overcome the barrier of highly imbalanced positive and negative samples in actual industrial data collection, a Balanced Defect Synthesis (BDS) data augmentation strategy is introduced to effectively enrich the morphological diversity of tiny defects. Secondly, a Wavelet Transform Convolution (WTConv) module is collaboratively integrated into the feature extraction network to expand the receptive field while preserving the high-frequency edge details of hairline cracks. Thirdly, a Group CBAM Enhancer (GCE) module is introduced to filter out high-reflection and brushed background noise through grouped attention and weight re-calibration mechanisms. Finally, addressing the difficulty of pixel-level alignment for tiny defects, an α-IoU loss function is utilized to improve the high-precision segmentation and localization capabilities by dynamically adjusting the gradient distribution. Comprehensive evaluations are conducted on two real-world electrical commutator surface defect datasets: KolektorSDD2 and KolektorSDD. Experimental results show that on the KolektorSDD2 dataset, compared to the YOLOv11 baseline, the Mask mAP@50 of WG-YOLOv11 increases from 85.2% to 89.2%, and the stringent metric Mask mAP@50:95 improves from 52.7% to 56.9%. Additional computational analysis on the same dataset validates that the proposed method maintains high efficiency, matching the baseline computational cost without compromising real-time inference speed. Furthermore, evaluations on the public MSD dataset confirm the model’s cross-domain generalization capabilities. The proposed framework effectively achieves a balance between detection accuracy, anti-interference robustness, and a lightweight architecture.

Algorithms doi: 10.3390/a19050421

Authors: Zhe Chen Fan Shi Jiawang Li Guangnian Li

With the development of shipping to harsh marine environment, it is very important to understand the transient behavior of a marine diesel engine in high sea conditions. Wave-induced hull motion will lead to severe load fluctuations and air-fuel ratio imbalance. In this study, an integrated simulation platform coupled with environmental loads, hull dynamics, propeller characteristics and a high-fidelity thermodynamic engine model was constructed to explore the response characteristics of the propulsion system. The model integrates a zero-dimensional multi-zone combustion method, turbocharger dynamic characteristics and an incremental PID governor, and has been verified based on the bench test data of TBD234V12 diesel engine and the 20 m Wigley standard ship. The simulation results under the sea conditions from level 7 to 9 show that the transient load has a nonlinear amplification effect. Specifically, from sea state 7 to sea state 9, the engine load fluctuation range expands by 2.0 times, while the main peak amplitude of speed fluctuation increases by 3.7 times. Furthermore, the peak exhaust pressure rises by 1.8 times, and the exhaust temperature fluctuation amplitude broadens by 35%. Frequency domain analysis further identified the low-frequency energy concentration phenomenon in the exhaust pressure spectrum and the precursor characteristics of compressor surge. The research results quantify the deterioration law of thermodynamic stability and mechanical stress under wave disturbance, and provide an important reference for the formulation of an engine robust control strategy and fatigue life assessment under high sea conditions.

Algorithms doi: 10.3390/a19050420

Authors: Vladimir Ternovski

Whether learned cognition can affect evolutionary outcomes remains a long-standing question. This study addresses a narrower mechanism: whether a model-based planner benefits from learned rules that explicitly condition on the action just taken. The testbed is a spatial artificial ecology with plants, shelters, a predator, reproduction, and a day/night cycle. Five rule-use arms are evaluated on matched simulation seeds. At age 200, agents switch to a weaker learned-lite planner that relies more strongly on learned rule predictions. The pre-specified hypothesis is that access to filtered action-conditioned rules improves outcomes relative to an otherwise identical no-rule-policy baseline, in which rules are still induced and stored but are not used for action selection. In thirty paired replicates under the default reproductive gates, the action-conditioned arm outperforms the no-rule baseline on all four pre-specified primary endpoints. The strongest effect is behavioural: the action arm produces 91.4 additional successful post-switch eating events per run (dz=1.56, 93.3% paired win rate, p<10−4). It also produces 10 additional crystallized clean-causal rules per replicate (dz=0.58, pt=0.0034). All four primary paired-tp-values remain significant after Bonferroni correction across the four-endpoint family. A diagnostic check shows that omitting reproductive cooldown from the planner’s rollout reverses the arm ordering on the same paired seeds; reinstating cooldown recovers the reported result. Two exploratory checks delimit the claim: broad unfiltered rule access can impair foraging, and a means–ends extension shifts behaviour toward reproduction without producing a robust whole-life fitness gain. Within this simulation, access to action-conditioned rules has a measurable effect on post-switch behaviour that is distinct from passive environmental prediction and from clean-crystallized rules alone.

Algorithms doi: 10.3390/a19050419

Authors: Yuto Tanimoto Kenji Fukumizu

While many multi-armed bandit algorithms assume that rewards for all arms are constant across rounds, this assumption does not hold in many real-world scenarios. This paper considers the setting of recovering bandits, where the reward depends on the number of rounds elapsed since the last time an arm was pulled. We propose a new reinforcement learning (RL) algorithm tailored to this setting, named the State-Separated SARSA (SS-SARSA) algorithm, which treats the elapsed rounds as states. The SS-SARSA algorithm achieves efficient learning by reducing the number of state combinations required for Q-learning/SARSA, which often suffers from combinatorial explosion for large-scale RL problems. Additionally, it makes minimal assumptions about the reward structure and has lower computational complexity. Furthermore, we prove asymptotic convergence to an optimal policy under mild assumptions. Simulation studies demonstrate the superior performance of our algorithm across various settings.

Algorithms doi: 10.3390/a19050418

Authors: Khaoula Nermine Khallouf Habib Benbouhenni Nicu Bizon

The intermittent nature of renewable power sources, nonlinear load effects, and harmonic distortions induced by power electronic converters complicate the maintenance of high energy quality in microgrid-connected hybrid renewable power systems. In a range of operating conditions, conventional strategies-including fractional-order proportional-integral (FOPI) controllers-frequently prove ineffective in delivering both robust harmonic mitigation and expeditious dynamic response. To surmount these constraints, the present paper puts forth an intelligent control solution that is predicated on a fractional-order fuzzy logic (FOFL). The FOFL is integrated into a multi-converter HRPS, comprising a photovoltaic generator, a lithium-ion battery power storage system, and a wind turbine equipped with a permanent magnet synchronous generator. A multifunctional voltage source inverter has been developed to control these parts, which are interfaced via a common DC bus. Through the implementation of MATLAB 2021 simulation studies, the efficacy of the suggested algorithm is verified and evaluated in comparison to the FOPI. The findings indicate that the FOFL enhances system efficacy by minimizing harmonic distortion, improving energy quality, and achieving a faster dynamic response under various circumstances. In the context of grid-connected microgrid environments, the FOFL has been demonstrated to offer superior overall energy management, robustness, and adaptability when compared to other evaluated strategies.

Algorithms doi: 10.3390/a19050417

Authors: Ryan V. Cooper Arslan Munir

As modern networks expand, the volume and destructiveness of cyberattacks continue to escalate, necessitating effective defense mechanisms. Intrusion Detection Systems (IDSs) are critical for maintaining network security; however, traditional signature-based systems often fail to detect zero-day attacks. This study explores recent advancements in Deep Learning (DL) for cybersecurity by analyzing and replicating the “IDS-ML” framework, an open-source repository for IDS development. We evaluate the performance of five deep learning Convolutional Neural Network (CNN) architectures adapted for intrusion detection via transfer learning on the CICIDS2017 dataset, and propose an enhancement by integrating Automated Machine Learning (AutoML) techniques that achieves a 94.7% reduction in model parameters while maintaining comparable accuracy, thus making our enhanced models suitable for deployment on edge devices. We further validate deployment feasibility by benchmarking both the baseline InceptionV3 and AutoML models on a Raspberry Pi 4, demonstrating an 18.7× inference speedup and 3.5× CPU reduction, with no change in predicted classes from model conversion. Our results confirm that lightweight AutoML architectures enable practical “zero-touch” edge-based intrusion detection on resource-constrained hardware.

Algorithms doi: 10.3390/a19050416

Authors: Suryakant Tyagi Sándor Szénási

Finding emotions in human speech is a difficult task. It is even harder for sounds without words, like laughs, gasps, and sighs. Normal audio models fail at this task because these sounds are very short and the audio patterns are complex. To fix this problem, we created a new model called the Hybrid Semantic-Acoustic Transformer. Our system uses a Wav2Vec 2.0 model to get acoustic features. At the same time, it uses a Whisper ASR model to get phonetic features. We mix these two types of data together using a Cross-Attention layer. We tested our model on the EmoGator dataset. This dataset has 32,130 audio files across 30 different emotion classes. We split the data strictly into 80% for training, 10% for validation, and 10% for testing. Our new model achieved an overall accuracy of 74.8%. We also did an ablation study. This study proves that using cross-attention is much better than simply adding the features together. Our final result is a 6.4% increase in the F1-score compared to the original EmoGator baseline model. This sets a new high score for classifying non-speech sounds in different noisy environments. Our model also reached over 90% precision when telling the difference between a ‘Sigh’ and a ‘Gasp’. Standard speech models usually fail at this specific task.

Algorithms doi: 10.3390/a19050415

Authors: Ádám Francuz Tamás Bányai

The challenge of inventory optimization is extremely important for all manufacturing companies, as inventory costs significantly impact operational efficiency. The Economic Order Quantity (EOQ) model was developed to address this issue, and it is widely used to formulate it, as it generally considers only a few parameters and a single objective. This research develops a simulation-based framework that integrates multiple EOQ-based inventory policies and performs multi-objective optimization using the NSGA-II algorithm. The framework optimizes total cost, fill rate, and average inventory level and finally generates a Pareto front as a result. To reduce computational costs, we use a machine learning-based random forest model, which replaces a significant amount of the simulations with predictions. This reduces the simulation cost to approximately one-sixth of the original, while the quality of the simulation changes only minimally, as the hypervolume value decreases by only 4%. The proposed framework can be used as an effective decision-support tool for inventory optimization under stochastic demand conditions.

Algorithms doi: 10.3390/a19050414

Authors: Hetali Tank Zhen Cao Juan Shan Ming Zhang

Osteoarthritis (OA) severity grading from hand distal interphalangeal (DIP) joint radiographs using the Kellgren–Lawrence (KL) scale is challenged by severe class imbalance, with higher grades (KL3 and KL4) markedly underrepresented in clinical datasets. To address this limitation, we propose a VGG19-based classification framework that systematically evaluates six training strategies targeting imbalance at the data level, algorithmic level, or in combination. Synthetic images for minority classes were generated using CycleGAN and subsequently filtered through rheumatologist validation. The evaluated strategies include baseline training, rheumatologist-validated synthetic augmentation (SD), oversampling (OS), focal loss (FL) optimization, and multiple combinations of these approaches. The results show that strategies incorporating oversampling demonstrated the most consistent and statistically robust improvements in minority-class performance. Specifically, the combination of synthetic data and oversampling (SD + OS) achieved the highest binary OA sensitivity (96.12%) and significantly improved OA F1 score compared to baseline (0.613 vs. 0.416, p = 0.029). The full combined strategy (SD + OS + FL) yielded the highest KL3 F1 score (0.527 vs. 0.280 baseline, p = 0.048) and significantly improved KL4 F1 score (0.730 vs. 0.570 baseline, p = 0.150). Importantly, all strategies maintained higher or similar overall performance with no significant change in majority-class performance (p > 0.10), indicating that improvements in minority classes were not achieved at the expense of sacrificing majority classes or overall model reliability. These findings suggest that the proposed imbalance-mitigation strategies may improve minority class OA detection, particularly when oversampling and validated synthetic augmentation are combined. It is worth noting that the above results are derived from a held-out test set comprising 1626 samples, among which only 43 are OA-positive due to data imbalance. The results should be treated as preliminary findings subject to change upon validation in larger cohorts of OA patients.

Algorithms doi: 10.3390/a19050413

Authors: Seyma N. Yildirim-Erbasli Okan Bulut

This study presents a controlled empirical and comparative analysis of existing data augmentation techniques for text generation in Turkish, a morphologically rich, low-resource language. A collection of 265 Turkish reading passages for Grades 4 and 5 was augmented using four techniques: paraphrasing with GPT-3.5-turbo (Generative Pre-trained Transformer 3.5 Turbo), back translation (Turkish–English–Turkish and Turkish–French–Turkish) via Google Translate, synonym replacement via GPT-3.5-turbo, and random insertion via GPT-3.5-turbo. Human evaluators assessed the fluency, coherence, grammaticality, logical flow, and naturalness of the augmented datasets. Each augmented dataset, along with the original, was then used to fine-tune a Turkish GPT-2-medium model, which was evaluated using automatic metrics such as BLEU (Bilingual Evaluation Understudy), ROUGE (Recall-Oriented Understudy for Gisting Evaluation), METEOR (Metric for Evaluation of Translation with Explicit ORdering), chrF (CHaRacter-level F-score), BERTScore (Bidirectional Encoder Representations from Transformers Score), and cosine similarity. According to the human evaluation of the original and augmented datasets, the original texts received the highest ratings, followed by those generated through random insertion, paraphrasing, synonym replacement, and back translation variants, with cosine similarity results between original and augmented texts showing a comparable trend; however, the differences between methods were generally small. The results from text generation indicate that models trained on the original dataset generally achieved slightly higher performance across evaluation metrics compared to those trained on augmented datasets. Among the augmented methods, synonym replacement showed marginally better performance, followed by back translation, random insertion, and paraphrasing; however, the differences between methods were small and not statistically significant.

Algorithms doi: 10.3390/a19050412

Authors: Tom Mashiach Gilad Katz Meir Kalech

In software engineering, detecting and managing code smells are pivotal for maintaining software quality and reducing the risk of defects. Code smells signify potential issues in code that, while not problematic in themselves, may indicate deeper design flaws or future complications. Traditional code smells detection methods, which compare code metrics against fixed or statistically derived thresholds, may not always yield the most accurate code smells relevant to specific software practices. Addressing this gap, this research introduces an innovative methodology that utilizes a neural threshold generator, trained via a cooperative critic, to dynamically generate threshold values for detecting code smells in software components. Although the critic is conceptually related to the discriminator in a Generative Adversarial Network (GAN), its training objective is aligned with rather than adversarial to that of the generator. By integrating relevant code metrics, the proposed model generates customized thresholds for each software component. Our current evaluation focuses on a set of 11 class-level code smells defined by single or AND-connected conditions. It then uses these thresholds to identify code smells, which serve as input features to train a defect prediction model. A key feature of our approach is a cooperative-critic feedback mechanism that continuously refines the thresholds based on the defect prediction outcomes, ensuring the model’s effectiveness in identifying potential software issues is consistently improved. This advanced approach has demonstrated superior defect prediction performance, as evidenced by improved metrics such as the F1-score, AUC-ROC, and AUC-PRC, compared with the results of a defect prediction model that uses the traditional thresholds. Our study underscores the effectiveness of generating context-specific thresholds through neural networks, suggesting a promising avenue for exploring related software practices.

Algorithms doi: 10.3390/a19050411

Authors: Zimeng Zhou Qingxiang Zhang Yanpeng Hu Fengwei Jing Jin Guo

Addressing the challenge of effectively detecting data tampering attacks in cyber-physical systems, this paper proposes an attack detection method based on prior information for the identification of a class of Hammerstein nonlinear systems measured by binary sensors. This method leverages the periodic structure of the system inputs and the statistical properties of the binary observation data to characterize the asymptotic properties of the parameter estimators; furthermore, by incorporating prior information regarding the system parameters, it constructs a detection criterion that enables the effective identification of attack behaviors. To enhance the computational efficiency of the algorithm in practical applications, a Multilayer Perceptron (MLP) is employed to approximate the implicit nonlinear inverse mapping, thereby circumventing the numerical difficulties associated with directly solving systems of nonlinear equations. On a theoretical level, the asymptotic distributions of the detection algorithm’s false alarm rate and missed detection rate are derived, and a systematic analysis is conducted on how detection performance is affected by factors such as system input period, prior information scope, and data length. Numerical simulations validate the efficacy of the proposed method; the results demonstrate that as the data length increases, both the false alarm rate and the missed detection rate of the algorithm decrease. Moreover, a broader scope of prior information leads to a lower false alarm rate but a higher missed detection rate, thereby illustrating the “double-edged sword” effect of prior information in the context of attack detection. This study provides a theoretical foundation and technical support for attack detection in nonlinear systems operating under conditions of data constraints and security threats.

Algorithms doi: 10.3390/a19050410

Authors: Yucong Ma Huiying Hou Xuerui Gan Zisu Zhao

With digital assets increasingly comprising a significant portion of personal wealth, the secure management and transfer of digital legacies have emerged as a pressing concern. Secret sharing offers a solution to this problem. However, distributing shares containing the unique private key for digital assets poses significant risks of theft or tampering, potentially leading to the illegal appropriation of user assets. This paper presents a leakage-resistant digital inheritance distribution scheme based on sparse-matrix secret sharing. It employs an efficient thresholding scheme that uses sparse matrices, achieving near-linear complexity for share reconstruction via a random striped matrix. Reconstruction time is significantly reduced compared to traditional polynomial interpolation methods. To address the realistic scenario where an asset owner holds multiple independent digital accounts, we propose a multi-account blinding and aggregation mechanism. This mechanism allows the dealer to establish isolated group keys for each account in a single round of communication, while preventing adversaries from linking different accounts to the same owner. A key-derivation and encrypted-transmission mechanism is then designed based on the aggregated group keys. Group keys are established by consensus among heirs, from which each heir derives a unique session key. Authenticated encryption ensures the confidentiality, integrity, and identity-bound transmission of shares. Through security proofs and experimental performance evaluation, it is demonstrated that the proposed scheme satisfies adaptive security requirements with the hash function H modeled as a random oracle, while all other cryptographic primitives (PRF, AES-GCM, HMAC) are assumed to be secure under standard computational assumptions.

Algorithms doi: 10.3390/a19050409

Authors: Kuhyun Lee Jihoon Hong Beom-Seok Kim Yuna Song Dong-Hee Lee

Object detection is a core component of industrial vision systems in manufacturing, infrastructure monitoring, and safety-critical sensing. While the mean average precision (mAP) averages the performance over all confidence thresholds, real-world deployment demands committing to a single operating threshold under score imprecision, distribution shifts, and asymmetric—often only approximately known—error costs. From a soft-computing perspective, deployment should explicitly manage this uncertainty rather than rely on a static validation optimum. We propose domain-specific and robust localization recall precision (DSR-LRP), a three-phase decision-support framework. The framework elicits soft domain preferences—such as asymmetric error costs, tolerable localization imprecision, and expected perturbations—from practitioner knowledge and encodes them as three quantitative parameters (k, αIoU, β). A cost-sensitive, threshold-local objective aggregates the performance within a robustness band around each candidate threshold, jointly capturing the accuracy and local stability. Finally, it yields an interpretable recommendation package comprising the operating threshold, its DSR-LRP score, and visual evidence. Experiments on four practical datasets (blood cell screening, wildfire smoke monitoring, pothole detection, and semiconductor sensor inspection) showed that DSR-LRP consistently selected operating thresholds that were robust and cost-aligned. For example, in pothole detection, an LRP-optimal threshold degraded by 15.6% under simulated shifts, while the DSR-LRP recommendation changed by only 1.8%. DSR-LRP complements global metrics such as the mAP and provides a soft-computing-oriented tool for reliable, evidence-driven deployment of industrial object detectors.

Algorithms doi: 10.3390/a19050408

Authors: Yantao Zhang Yangyang Li Shuxin Wang Jingxuan Wang Robail Yasrab Xinli Wu

Accurate prediction of invasive species diffusion is essential for effective management and ecological conservation. Existing spatio-temporal Cox process models face limitations due to the separability assumption, which fails to capture spatio-temporal coupling dynamics inherent in biological diffusion processes. This study proposes a Spatio-Temporal Interaction Kernel Cox (STIK-Cox) model that constructs a non-separable conditional intensity function integrating baseline intensity, spatial and temporal proximity kernels, seasonal fluctuation, and a spatio-temporal interaction term. The model employs maximum likelihood estimation with Limited-memory Broyden–Fletcher–Goldfarb–Shanno with Bounds (L-BFGS-B) optimisation and incorporates SHapley Additive exPlanations (SHAP) for interpretability analysis. Using the Vespa mandarinia (Hymenoptera, Vespidae) monitoring dataset from Washington State, the model achieves a comprehensive accuracy score of 0.957, a capture rate of 98.74% at a 0.5° threshold, and a mean prediction error of 0.0802°. K-function analysis confirms effective capture of spatial clustering patterns, while SHAP analysis reveals longitude as the primary predictive driver. The non-separable design outperforms conventional methods including inverse distance weighting and Poisson point processes. This framework demonstrates the potential of non-separable spatio-temporal point processes for invasive species early warning, providing a scientific basis for targeted monitoring and resource allocation in ecological management.

Algorithms doi: 10.3390/a19050407

Authors: Qiang Huang

Few-Shot Class-Incremental Learning (FSCIL) aims to learn new classes with only a few samples, making it more challenging than traditional Class-Incremental Learning (CIL) due to the scarcity of available samples. The imbalance in sample distribution further complicates balancing the abundant base data with the scarce incremental data. While the model must fully leverage the extensive base data to guide the learning of subsequent tasks, it must also avoid over-relying on these data, as doing so could degrade its generalization capability and impede the learning of new incremental tasks. To address these challenges, we propose a novel framework for few-shot incremental learning, incorporating tailored prompt alignment strategies for both the base and incremental session. In the base session, we strike a balance between task-specific and task-agnostic knowledge to preserve the model’s generalization ability. In the incremental session, we mitigate the overfitting issue typically associated with few-shot learning. Furthermore, to tackle the prototype network bias caused by the imbalance in sample distribution, we propose a subspace prototype aggregation module, which effectively alleviates prediction bias in the incremental phase. Extensive experiments conducted on three benchmark datasets—CIFAR-100, miniImageNet, and CUB-200—demonstrate that our approach achieves state-of-the-art (SOTA) performance in FSCIL.

Algorithms doi: 10.3390/a19050406

Authors: Anesito Cutillas Fritz Bacalso Christine Joy Tomol Melanie Albarracin Rose Ann Campita Eingilbert Benolirao Kafferine Yamagishi Lanndon Ocampo

Despite advances in using multi-criteria decision-making (MCDM) methods and their fuzzy set extensions for human evaluations of large language models (LLMs), several gaps remain in the literature, particularly in task-specific evaluations that offer a more tractable and interpretable approach. Thus, this work develops a generalized intuitionistic fuzzy MCDM approach that bridges methodological gaps by outlining two contributions. First, the integration of SWARA (Stepwise Weight Assessment Ratio Analysis) and WINGS (Weighted Influence Non-linear Gauge System) is demonstrated to compute the priority weights of the evaluation criteria, thereby augmenting the independence limitation in prior relevant studies. Second, we introduce a newly improved IF-EDAS (intuitionistic fuzzy Evaluation based on Distance from Average Solution) that preserves more uncertain information and provides a more natural extension of the canonical EDAS framework, starting with the adoption of the IFWAM (intuitionistic fuzzy weighted arithmetic mean) operator for a more intuitive approach in generating the intuitionistic fuzzy average solution vector. Also, the proposed IF-EDAS variant employs three decision rules and the Hamming distance metric in its novel computational approach. The proposed hybrid approach was deployed in two case studies evaluating five popular LLMs for text summarization across seven interdependent criteria. Results show that SWARA initially prioritizes accuracy, coherence, and consistency, but these were revised when accounting for criteria interdependence, with coherence and language quality emerging as the most preferred criteria. Both case studies suggest that Gemini may perform favorably, while Copilot may consistently rank last. The findings of the case studies share similar insights with those of three other similar IF-EDAS variants, although our claims may have limited external validity, which requires more case studies and experts in future task-specific human evaluations. The proposed approach, along with its deployment in two case studies, demonstrates human evaluations of LLMs with greater computational interpretability, which contribute to the general MCDM literature.

Algorithms doi: 10.3390/a19050404

Authors: Aveer Nannoolal Andries P. Engelbrecht

This study provides a comprehensive evaluation of bagging ensemble models for financial time series (FTS) classification and addresses a gap in the literature regarding how bootstrap methods, ensemble sizes, voting mechanisms, and loss functions jointly influence model performance. The analysis evaluates decision tree (DT), logistic regression (LR), and multi-layer perceptron (MLP) ensemble models modified by six time series bootstrap methods, five ensemble sizes, and three voting mechanisms across six FTS data sets. The study also examines the influence of entropy- and profit-based loss functions within particle swarm (PSO) and quantum-inspired particle swarm (QPSO) optimization for weighted voting. The results show that LR-based ensembles provide the strongest overall performance and outperform ARIMA, DT, LR, MLP, and LSTM baseline models on both accuracy and profit metrics. Bootstrap effects are model specific. DT and MLP ensembles perform best under the Tukey bootstrap, while LR ensembles achieve strong results under the block bootstrap, the sub-sample bootstrap method, and the Tukey method, and remain the strongest performers across all bootstrap configurations. Optimized voting mechanisms yield clear improvements over equal-weight majority voting, with the profit loss function producing the most consistent gains. The findings also indicate that FTS classification problems exhibit an optimal range of ensemble sizes, as larger ensembles do not always improve performance. The study contributes a systematic assessment of ensemble design choices for FTS classification and highlights the importance of jointly considering bootstrap diversity, ensemble size, and voting strategy when developing ensemble models for financial applications.

Algorithms doi: 10.3390/a19050405

Authors: Mohammad Shehab Afaf Edinat Mariam Al Ghamri Mamdouh Gomaa Fatima Alhaj Israa Wahbi Kamal Ahmed E. Fakhry

Machine learning (ML) has become a key enabling technology for optimizing renewable energy systems and supporting global sustainability objectives. This paper presents a comprehensive review of recent advances in ML-based optimization techniques applied to clean and renewable energy systems, with particular emphasis on wind energy, hybrid energy systems, energy storage, and intelligent energy management. A systematic literature review covering peer-reviewed publications from 2021 to 2025 was conducted, resulting in the analysis of 138 high-quality journal and conference studies. The reviewed studies were categorized according to evolutionary algorithm-based hybrid models, classical neural networks, and deep learning architectures, including Convolutional Neural Network (CNN), LSTMs, GRUs, and attention-based models. The analysis demonstrates that hybrid ML–metaheuristic frameworks significantly enhance forecasting accuracy, system reliability, fault diagnosis, and multi-objective optimization compared to traditional methods. These intelligent approaches directly contribute to Sustainable Development Goals SDG-7 (Affordable and Clean Energy), SDG-9 (Industry, Innovation, and Infrastructure), and SDG-13 (Climate Action). Key challenges and future research directions are discussed, highlighting the need for scalable, explainable, and real-time ML solutions to enable resilient, low-carbon, and sustainable energy systems.

Algorithms doi: 10.3390/a19050403

Authors: Changhui Han Mengmeng Zhang Jie Zhang Xiaolong Ma

The joint optimization of routing and three-dimensional loading is a highly complex NP-hard combinatorial problem, particularly when stringent center-of-gravity (CoG) stability constraints are required for unmanned aerial vehicle (UAV) operations. Existing algorithms typically adopt a route-first, load-second evaluation strategy for these interconnected components, often yielding distance-optimal yet physically infeasible solutions. To address this bottleneck, this paper formulates the Three-Dimensional Loading-Constrained UAV Routing Problem (3DLC-UAVRP), integrating unloading sequence consistency, spatial packing feasibility, and CoG deviation control into the routing decision process. A guided collaborative optimization framework, GLS-WSCPA, is proposed, coupling an Improved White Shark Optimization (IWSO) algorithm for global route exploration with a Human-like Divide-and-Conquer Packing Strategy (HLDCPS) for spatial arrangement. Unlike conventional decoupled approaches that treat loading feasibility as a post hoc filter, a Center-of-Gravity-Guided Path Adjustment (CGPA) and Local Loading Repair (LLR) mechanism is introduced to establish a dynamic feedback loop between routing search and loading evaluation, so that CoG violations are actively translated into guided routing perturbations rather than simply triggering solution rejection. Experimental results demonstrate that GLS-WSCPA generally achieves better solutions than the compared algorithms across the tested problem scales, with the performance gap tending to widen as the instance size increases within the tested range. Ablation studies verify the complementary roles of CGPA and LLR, and sensitivity analysis confirms that moderately relaxing payload and CoG constraints reduces routing distance within safety boundaries. Case analysis shows that the proposed method reduces fleet size by 20% and total delivery distance by 6.85% compared to traditional decoupled strategies.

Algorithms doi: 10.3390/a19050401

Authors: Wenrong Song Yuan Xu Bin Lyu Wenbin Liu Yuxuan Zhang Jin Wang

Oil enterprises face the challenge of reconciling escalating energy conservation targets with persistent production requirements, necessitating sophisticated electricity management solutions. The conventional ton-per-kWh allocation approach, often manually adjusted based on historical production and planning data, lacks a scientific basis and fails to accurately identify efficiency differences or assess energy-saving potential, making it difficult to convince participating units. To address this, we propose a dynamic spatiotemporal allocation scheme and develop a multi-objective optimization model that integrates electricity efficiency, operational stability, and production priority. The model incorporates nonlinear efficiency terms, stability components, and priority-weighted items, with constraints including total balance, monthly adjustment limits, and key area protection. Central to the efficiency term is the accurate prediction of liquid production from electricity consumption. We decompose electricity use into three components—core production electricity, auxiliary production electricity, and product transportation electricity—and derive their proportional coefficients through regression of historical data, enabling high-precision liquid production prediction via machine learning using the Light Gradient Boosting Machine (LGBM). The resulting constrained optimization problem is solved using the Sequential Least Squares Programming (SLSQP) algorithm. Validation using both simulated data and Daqing Oilfield field data demonstrates that the scheme effectively achieves electricity reduction targets while significantly mitigating associated liquid production loss, reducing it by 18.0% in simulated experiments and 32.5% in field validation compared to the conventional ton-per-kWh method. This offers a scientific and adaptive electricity management framework that supports refined energy control and facilitates the petroleum industry’s green and low-carbon transformation.

Algorithms doi: 10.3390/a19050402

Authors: Alexander Khanov Maksim Gozhan Denis Butusov Yulia Bobrova Valerii Ostrovskii

Cycle-to-cycle variability of switching parameters inherent to memristive devices introduces significant problems in the design of neuromorphic systems and non-volatile memory. This study investigates the dynamics of a second-order memristive system incorporating capacitive effects that model parasitic charge within individual memristors, addressing both the technical need for accurate analysis of complex regimes and the demand for exploratory environments. Simulations were performed using CUDAynamics, an interactive software suite developed by the authors, which utilizes parallel computing, primarily via NVIDIA Compute Unified Device Architecture (CUDA). It integrates multiple analysis tools for dynamical systems, including bifurcation diagrams, the largest Lyapunov exponent and periodicity mapping, and interactive navigation in multidimensional parameter spaces. The memristive system was discretized applying multiple integration methods with a fixed time step and various waveforms of the input signal. Analysis tools revealed well-defined regions of chaotic dynamics in the memristor resistance parameter space as functions of input signal properties. Sinusoidal and triangular waveforms produced topologically similar distributions of dynamical regimes, whereas the square waveform, mimicking digital inputs, generated distinct dynamical patterns while still preserving chaotic trajectories under specific conditions. Interactive visualization capabilities of CUDAynamics effectively demonstrate attractor evolution and hysteresis deformation, providing immediate visual feedback that significantly enhances conceptual comprehension of nonlinear feedback mechanisms. Beyond its practical implications for the design of analog and digital memristive devices, CUDAynamics offers a scalable, open-source toolkit to aid researchers and engineers in exploring complex dynamical phenomena.

Algorithms doi: 10.3390/a19050400

Authors: Yanxin Wang Nwabueze Emekwuru Chengqian Jin Fernando Auat Cheein

This study proposes an intelligent spraying vehicle speed control system integrating real-time canopy density detection with a fuzzy PID control algorithm. Utilizing LiDAR-acquired 3D point cloud data for canopy density calculation, the system dynamically adjusts PID parameters through fuzzy logic to achieve coordinated optimization of vehicle speed and spray volume. Based on the designed canopy density prediction model, a MATLAB/Simulink co-simulation framework integrating canopy perception with vehicle dynamics was established. Simulation results based on the MATLAB/Simulink platform demonstrate that the fuzzy PID controller achieves superior performance compared to conventional PID control. While maintaining a tracking accuracy of ±0.15 m/s, the proposed controller reduces speed overshoot by 5.8 percentage points. The developed control system ensures optimal speed tracking under varying canopy conditions, providing an extensible technical framework for intelligent sprayer vehicles.

Algorithms doi: 10.3390/a19050398

Authors: Samin Semsar Kiran Laxmikant Prabhu Gabriella Waters James Foulds

There are ongoing discussions about predictive policing systems being unfair, for example, by exhibiting racial bias. Law enforcement in some cities, such as Los Angeles, California, and Baltimore, Maryland, have initiated the integration of these systems into their decision-making processes, and some of these systems were advertised as being unbiased. However, later studies discovered that these methods could also be unfair due to feedback loops and being trained on historically biased recorded data. Comparative studies on predictive policing systems are few and insufficiently comprehensive. Crucially, the relative fairness of predictive policing methods with regard to traditional hot spot-based policing has not been established. Moreover, the relationship between fairness and accuracy is complex and requires further study. Furthermore, the case of Baltimore City, Maryland, USA, has not yet been systematically analyzed despite its relevance as an early adopter of predictive policing technologies with a fraught history of social justice concerns around policing. An improved understanding of these questions could better inform policy decisions around predictive policing technologies both in Baltimore and beyond. Therefore, in this work we perform a comprehensive comparative simulation study on the fairness and accuracy of predictive policing technologies in Baltimore. Our results suggest that the situation around bias in predictive policing is more complex than previously assumed. While we find that predictive policing exhibits bias due to feedback loops, as previously reported, we also find traditional hot spot-based policing to have similar issues. Although predictive policing is found to be more fair and accurate than hot spot policing in the short term, it also amplifies bias more quickly, suggesting the potential for worse long-run behavior. In Baltimore, the bias in these systems tended toward over-policing White neighborhoods in some cases, unlike in previous studies. However, when the analysis was restricted to some specific crime types, this tendency differed. Overall, this work demonstrates a methodology for city-specific evaluation and compares behavioral tendencies of predictive policing systems, showing how such simulations can reveal inequities and long-term tendencies. We recommend that authorities and community stakeholders use simulation methodologies to assist in collaboratively navigating the complexities around fairness in predictive policing.

Algorithms doi: 10.3390/a19050399

Authors: Meryem Cherrate Imane El Manaa My Abdelouahed Sabri Yassine Abouch Ali Yahyaouy Abdellah Aarab

Designing effective temporal modeling strategies for video-based sign language recognition (SLR) remains challenging, particularly in low-resource settings where the behavior of modern architectures is not fully understood. In this study, we present a controlled comparative evaluation of temporal models, including recurrent architectures (RNN, LSTM, GRU) and a Transformer encoder, within a unified spatio-temporal framework based on a shared MobileNetV2 feature extractor. All models are trained and evaluated under identical conditions on a curated subset of the WLASL dataset (37 classes), ensuring a fair and reproducible comparison. The results show that recurrent models consistently achieve higher performance than the Transformer-based approach in data-constrained scenarios, with the CNN–LSTM model reaching an accuracy of 90.02%. In contrast, the Transformer model exhibits lower generalization capability, which may be attributed to its higher data requirements. Additionally, increasing architectural complexity through hybrid temporal designs does not result in performance improvements. These findings suggest that simpler recurrent architectures remain effective for temporal modeling in limited data settings and highlight the importance of aligning model complexity with data availability for practical SLR applications.

Algorithms doi: 10.3390/a19050397

Authors: Bakhtgerey Sinchev Askar Sinchev Aksulu Mukhanova Tolkynai Sadykova Anel Auyezova Kuanysh Baimirov

This paper studies subset construction in NP-complete problems from the perspective of structured exploration of combinatorial search spaces. Classical approaches rely on exhaustive enumeration of subsets, which leads to exponential growth in time and memory requirements. To address this limitation, we introduce an indexed framework based on the correspondence between a finite set and its associated index set. Within this framework, subsets are represented as ordered index sequences, allowing subset construction to be reformulated as a constraint-guided search process over index space. Candidate subsets are characterized by numerical descriptors derived from their indices (referred to as index certificates), which guide and filter the construction process. Subset generation is further organized through admissible index intervals that restrict feasible transitions and reduce the effective search space. The framework is based on an index-based representation and structured traversal of pairwise index combinations. Computational experiments on representative instances illustrate the behavior of the indexed construction procedure and indicate its efficiency relative to classical enumeration-based methods for small and medium-sized instances. The proposed approach provides a structured perspective on combinatorial search and offers a basis for further development of algorithms based on constrained exploration of subset structures.

Algorithms doi: 10.3390/a19050396

Authors: Tian Meng Guoping Jiang

To address the challenges of difficult value quantification, lack of market benchmarks, and scarcity of historical data for embedded software amidst the intelligent transformation of equipment systems, this study develops a scientific price estimation method based on functional capability contribution. A nonlinear pricing model is constructed to accurately characterize the two-stage evolution of software price: diminishing marginal utility during the mature technology accumulation stage and exponential growth during the technical bottleneck breakthrough stage. To ensure the consistency of pricing logic between hardware and software, a penalty function is innovatively designed to modify the standard likelihood function, effectively transforming practical business logic into a model regularization term. Parameter estimation is achieved by employing a Bayesian inference framework integrated with operational constraints, utilizing Markov Chain Monte Carlo (MCMC) sampling to realize robust posterior inference under small-sample constraints. Empirical analysis demonstrates that the proposed method achieves superior cross-domain data transfer performance compared to traditional baseline models, with a Leave-One-Out Cross-Validation (LOOCV) Mean Absolute Percentage Error (MAPE) of 21.2%. This research provides a practical value-oriented price estimation method for embedded equipment software pricing.

Algorithms doi: 10.3390/a19050395

Authors: Aiman Moldagulova Zhuldyz Kalpeyeva Raissa Uskenbayeva Nurdaulet Tasmurzayev Bibars Amangeldy Yeldos Altay

Low-cost IoT-based air quality sensors enable dense monitoring networks but suffer from significant measurement noise and instability particularly in dynamic environments. Conventional fixed-window smoothing reduces noise but introduces a trade-off between signal stability and temporal responsiveness, often attenuating short-term pollution events. This paper proposes an adaptive filtering algorithm that dynamically adjusts the averaging window size based on short-term signal variability. The method relies on real-time variance estimation to balance noise suppression and sensitivity to rapid changes without increasing computational complexity. The approach is implemented within an IoT-based monitoring framework and evaluated using parallel measurements with a certified reference device. Comparative analysis against a certified reference device demonstrates strong agreement, with Pearson correlation coefficients reaching r = 0.88 for PM2.5 and r = 0.86 for PM10, and low error levels (RMSE ≈ 2.1–2.2 µg/m3). The proposed adaptive filtering approach preserves temporal dynamics while improving signal stability and robustness compared to raw and fixed-window filtering. In addition, this method improves event detection stability, achieving low false alarm rates and near real-time response (latency < 1 sampling interval), supporting RPA-based workflow triggering. The results show that the proposed adaptive filtering provides an efficient and lightweight solution for real-time signal processing on resource-constrained devices, making it suitable for large-scale deployment in environmental monitoring systems.

Algorithms doi: 10.3390/a19050394

Authors: Huthaifa Al-Omari Murad A. Yaghi Layan Alrifai

Accurate precipitation forecasting is essential for water resource management, flood early-warning systems, and agriculture, but remains difficult because of the nonlinear and highly variable spatiotemporal nature of rainfall. This paper compares four deep learning architectures—a standalone LSTM, a standalone CNN, a hybrid CNN–LSTM, and a Transformer encoder—against three classical baselines (persistence, day-of-year climatology, and per-grid-point ARIMA) for daily precipitation forecasting over Washington State at lead times of one to four days. A 40-year ERA5 dataset (1985–2024) of near-surface air temperature, mean sea-level pressure, and total precipitation is split into training (1985–2012), validation (2013–2015), and test (2016–2024) periods, with the test years held out completely. Each (model, horizon) is trained with three random seeds and evaluated in physical units (mm/day). On the held-out test period, the hybrid CNN–LSTM achieves the lowest RMSE at every horizon h≥2, with R2=0.576±0.007 and RMSE =15.08±0.07 mm/day at h=4. Diebold–Mariano tests, paired t-tests, and bootstrap 95% confidence intervals confirm that the CNN–LSTM advantage over the LSTM is statistically significant at horizons 2–4 (but not at h=1), while CNN–LSTM is significantly better than every classical baseline and the Transformer at every horizon. The headline result is reproduced under a rolling-origin temporal cross-validation across three non-overlapping splits (R2∈[0.576,0.590]). Practically, the sub-millisecond inference cost of the CNN–LSTM makes it directly deployable in operational forecasting pipelines used for flood early-warning, irrigation scheduling, and reservoir management, where even modest improvements in 3–4-day-ahead RMSE translate into measurable risk reduction and improved decision lead time for water managers and emergency planners.

Algorithms doi: 10.3390/a19050393

Authors: Yijie Zhang Jizhou Chen

To address the challenges of high computational complexity, inferior path performance, and the balance between path quality and efficiency in traditional 3D omnidirectional path planning algorithms for UAVs, this study proposes an innovative precision algorithm for solving 3D omnidirectional shortest paths. The algorithm innovatively introduces the concepts of circling path and overpass path, reducing three-dimensional omnidirectional path computation to two-dimensional processing. It designs a three-view obstacle detection algorithm to achieve efficient obstacle avoidance judgment, formulates separate path-solving strategies for discrete and continuous obstacles, respectively, and obtains optimal solutions through recursive adjustments and path optimization. Experimental results demonstrate that compared to A* and Theta* algorithms, our approach achieves shorter path lengths with superior stability; the proposed algorithm achieves a 21.86% reduction compared to RRT*, 10.48% compared to A*, and 0.89% compared to Lazy_Theta*. In addition, the proposed algorithm exhibits enhanced adaptability in high-obstacle environments (particularly irregular obstacles). These findings provide an effective solution for complex spatial path planning in UAV applications.

Algorithms doi: 10.3390/a19050391

Authors: Ndivhuwo Nelufhangani Daniel Maposa

This study investigates the effectiveness of Bayesian probabilistic methods for stock price forecasting on the Johannesburg Stock Exchange by implementing and comparing Gaussian process regression (GPR), Bayesian long short-term memory (Bayesian LSTM), and Bayesian neural networks (BNNs). Using daily open, high, low, close, and volume (OHLCV) data and engineered technical indicators for FirstRand and Discovery from January 2005 to June 2025 (5187 observations), models were trained and evaluated with the mean absolute error (MAE), root mean squared error (RMSE), and mean squared error (MSE). The GPR produced reliable, well-calibrated intervals in relatively stable regimes, but its performance degraded on the more volatile Discovery series. Bayesian LSTM delivered conservative uncertainty estimates with wide predictive intervals but showed the largest point forecast errors. The BNNs achieved the best balance between accuracy and uncertainty quantification, producing the lowest errors for FirstRand and competitive performance for Discovery. Comparative analysis indicates that BNNs are most suitable when point accuracy and calibrated uncertainty are both priorities, GPR is valuable for smaller or more stable data regimes, and Bayesian LSTM is preferable where conservative, risk-conscious intervals are required. This study highlights the practical value of embedding uncertainty into financial forecasts and recommends matching Bayesian model choice to market volatility, data availability, and decision maker risk appetite.

Algorithms doi: 10.3390/a19050390

Authors: Subhashree Mohapatra Jai Govind Singh Subham Pankaj Samantaray Manohar Mishra

In 1997, Long Short-Term Memory (LSTM) networks were proposed, which significantly changed the landscape of sequential data analysis by resolving the critical issue of the vanishing gradient problem in recurrent neural networks (RNNs). Over the last 25 years, LSTM has advanced from its inception as an innovative solution to its widespread adoption as an essential tool in various fields, including natural language processing (NLP), speech recognition, financial prediction, and healthcare analytics. The present study is a bibliometric review of the evolution of LSTMs. The evolution of LSTM is discussed in terms of its theoretical advancements, architectural developments, and its applications. The study is based on data obtained from the Scopus database, which is then analyzed to identify publication patterns, prominent authors, prominent institutions, and global contributions to the field. The present study is an insightful review of the evolution of LSTM, highlighting its developments and advancements, as well as its applications, to identify its future scope.

Algorithms doi: 10.3390/a19050392

Authors: Georgios Dimitrios Gkologkinas Konstantinos Ntouros Eftychios Protopapadakis Vasilis Drimzakas-Papadopoulos Nikolaos Samaras

Accurate field-scale wheat yield estimation is essential for precision agriculture, farm-level decision-making, and food security planning. However, operational studies conducted under real commercial farming conditions in the eastern Mediterranean remain limited. This study investigated whether multi-temporal Sentinel-2 imagery could support reliable wheat yield estimation across nine commercial wheat fields near Ptolemaida, Greece, during the 2023–2024 growing season. Both durum and common wheat fields were included, and combine-harvester yield maps were used as ground-truth observations. Six regression algorithms—the Random Forest (RF), Support Vector Regression (SVR), k-nearest neighbors (KNN), Decision Tree (DT), LASSO regression, and Gaussian Process Regression (GPR) algorithms—were evaluated using three feature configurations: raw Sentinel-2 spectral bands only (Sentinel-only (SO)), spectral bands combined with vegetation indices (Sentinel+Indices, SI), and vegetation indices only (Indices-only, IO). Model generalization was assessed through a strict Leave-One-Field-Out (LOFO) cross-validation protocol, and the method of SHapley Additive exPlanations (SHAP) was used to interpret model behavior and identify the most influential spectral regions and phenological stages. RF achieved the highest predictive accuracy, with a MAPE of 7.90% and an RMSE of 45.15 kg decare−1 under the SO configuration, demonstrating a statistically significant improvement over DT and KNN models (p<0.05). SHAP analysis indicated that model predictions were mainly driven by SWIR-1, NIR-narrow, and red-edge bands acquired during late grain filling and maturity, while vegetation indices contributed limited additional information. These findings suggest that raw multi-temporal Sentinel-2 spectral bands are highly effective for field-scale wheat yield estimation within the scope of this study, although further validation across diverse growing seasons and geographic regions is required to confirm broad operational sufficiency.

Algorithms doi: 10.3390/a19050389

Authors: Miguel Arcos-Argudo Rodolfo Bojorque David Galarza-García

Encrypted transport increasingly limits the visibility required by intrusion detection systems (IDS), motivating payload-free learning from flow statistics and protocol metadata. We introduce GCP, a graph-contrastive pretraining framework that casts flows as nodes in a sparse graph and learns transferable node embeddings via an InfoNCE-style objective with graph-specific augmentations. The learned encoder is evaluated through frozen-embedding linear probing and cross-dataset out-of-domain (OOD) transfer, within a fully scripted pipeline that freezes run manifests and artifacts to make every reported number traceable and reproducible. Experiments cover enterprise IDS and encrypted DNS/DoH traffic using CICIDS2017, UNSW-NB15, and DoH-Combined at three label granularities (L1/L2/L3), for both binary detection (y) and finer-grained targets (ymulti), aggregated over five fixed split seeds with 95% confidence intervals. Results show that GCP yields a pronounced in-domain advantage on UNSW-NB15 for y (Macro-F1 ≈0.993) while substantially reducing false-alarm rate (FAR ≈0.013) compared with strong tabular baselines. In feature-separable regimes (CICIDS2017 and DoH L1/L2), boosted-tree and supervised baselines remain difficult to surpass, but ablations confirm that graph structure alone is insufficient without contrastive pretraining. OOD transfer is strongly source–target dependent, with the most reliable transfer within closely related DoH domains, highlighting dataset shift as a first-class evaluation criterion for encrypted-traffic IDS.

Algorithms doi: 10.3390/a19050388

Authors: Garrett Rose Senthil Krishnamurthy

The development of offshore wind energy has increased the demand for reliable submarine transmission systems. In South Africa, research remains constrained due to the lack of operational offshore wind farms, despite favorable geographical conditions and persistent energy challenges such as load-shedding. Submarine cable faults, primarily caused by manufacturing deficiencies, environmental factors, and human activities, contribute significantly to system downtime while accounting for only a small portion of overall installation costs. This study reviews submarine cable fault identification, classification, pre-determination, and localization techniques. Conventional methods, including time-domain reflectometry, the Murray loop, the Varley loop, and impulse-based techniques, are reviewed alongside artificial neural network models, such as convolutional and deep learning architectures. Findings imply that traditional techniques offer low error margins but lack the accuracy needed for pinpointing exact faults, as faults may extend over several kilometers. In contrast, neural network-based methods, particularly when integrated with signal processing methods, significantly improve fault classification and localization accuracy. The study concludes that hybrid approaches combining conventional diagnostic techniques with neural networks offer a robust framework for submarine cable fault analysis, providing real-world solutions to enhance reliability and efficiency in future offshore wind transmission systems.

Algorithms doi: 10.3390/a19050387

Authors: Haijian Wu Xiaoguang Bao

In many real-world logistics systems, two-echelon distribution structures and clustered customer demands often coexist. However, traditional Two-Echelon Vehicle Routing Problems (2E-VRPs) mainly focus on the coordination between depots, satellites, and customers, while usually ignoring clustered customer service requirements. To fill this research gap, this study investigates a novel variant of the 2E-VRP, called the 2E-VRP with Clustered Customers (2E-VRP-CC). In this problem, customers in the second echelon are partitioned into predefined clusters, and all customers within a cluster must be visited consecutively by the same vehicle. For the problem, a Mixed-Integer Linear Programming (MILP) model is first established, followed by the development of an Adaptive Large Neighborhood Search (ALNS) algorithm integrated with a local search method. To validate the effectiveness of the proposed algorithm, comparisons are conducted on instance sets adapted from the literature. For the traditional 2E-VRP, which is a special case of the 2E-VRP-CC, the proposed algorithm is compared with existing methods in the literature. For the proposed 2E-VRP-CC, it is compared with the CPLEX solver. Extensive computational experiments demonstrate that the proposed algorithm can achieve high-quality solutions within relatively short computing times, confirming its effectiveness and efficiency. In addition, sensitivity analysis shows that the number of customer clusters has a significant impact on transportation costs. The results indicate that moderately increasing the number of customer clusters can effectively reduce operational costs and provide practical decision support for customer clustering design and two-echelon logistics planning.

Algorithms doi: 10.3390/a19050386

Authors: Neal N. Xiong Weiqing Long Dacheng He Xiangwei Meng Zulong Diao Sergey M. Avdoshin Yevgeni Koucheryavy

In the era of data-driven education, educational social networks generate large volumes of high-dimensional and complex-structured data through learner interactions, collaborative activities, and resource-sharing behaviors, posing significant challenges to traditional unsupervised learning methods. Such data often exhibit non-convex distributions, heterogeneity, and noise sensitivity, making conventional clustering approaches insufficient for capturing their intrinsic structural relationships. To address this issue, this paper proposes Quantum Fidelity-Based Graph K-Means (QGKM), a clustering framework for robust pattern recognition in educational social networks. Specifically, QGKM employs quantum state encoding to map complex educational data into a quantum state space and utilizes quantum fidelity as a similarity metric to uncover latent correlations that Euclidean distance cannot effectively capture. In addition, the incorporation of k-nearest neighbor graphs preserves the local geometric structure of learner interaction networks, while a deterministic greedy hierarchical merging strategy eliminates the instability caused by random initialization. Experimental results on seven real-world datasets demonstrate that QGKM consistently outperforms classical K-Means in clustering accuracy. The proposed framework provides an effective solution for learning pattern discovery, learner profiling, and intelligent recommendation in digital education environments.

Algorithms doi: 10.3390/a19050385

Authors: Ahmed K. Ali Mudhar A. Al-Obaidi Alhassan H. Ismail M. N. Mohammed Dhifaf Sadeq

Proton Exchange Membrane Fuel Cells (PEMFCs) are crucial for the advancement of environmentally friendly hydrogen cars. It is one of the promising solutions for alternatives to conventional engines, primarily due to their ability to convert hydrogen into electricity. Fuel cell systems have a complex and non-linear mathematical model. Accurate identification of unknown parameters of the PEMFC mathematical model is an important aspect in energy conversion. This research intends to provide a novel meta-heuristic algorithm, which is known as the Improved Dhole Optimization Algorithm (IDOA), to estimate the unknown parameters of PEMFC models. The proposed IDOA is inspired by the collective hunting behavior of dholes. In this algorithm, candidate solutions are systematically arranged and dynamically updated to enhance the overall search process. The objective function is to minimize the sum of squared errors (SSE) between the actual and model-estimated voltages obtained using the proposed IDOA algorithm. In this research, three commonly known PEMFC benchmark models, NedStack PS6, BCS 500 W, and Horizon 500 W, are utilized to assess the performance of the IDOA algorithm. Also, the obtained results are compared against each other to validate their effectiveness. The comparative performance with an array of known optimization algorithms reported in the literature indicates that the IDOA algorithm has low estimation error, an excellent convergence rate, and superior robustness. Furthermore, these results support the appropriateness of the proposed IDOA algorithm for high-accuracy PEMFC modeling in energy conversion models.

Algorithms doi: 10.3390/a19050384

Authors: Iosef Georgian Teșulă Adrian Zamfirel Nicolae Goga Răzvan Crăciunescu

The explosive growth of the Internet of Things (IoT) has expanded the attack surface across industrial systems, smart cities, healthcare, and homes, motivating a synthesis of recent advances in machine learning for IoT security and a clear statement of remaining gaps. This review conducted a systematic search of MDPI, IEEE Xplore, Nature, ScienceDirect, and SpringerLink for publications from 2023 to 2025, screening them for domain relevance and organizing findings into a taxonomy of ML methods, threat types, and deployment contexts, with particular attention to datasets, edge constraints, and privacy considerations. We find that the field is shifting from signature-based detection to supervised and deep learning approaches that report high accuracy on benchmark traffic, while federated learning enables privacy-preserving, distributed intrusion detection with near-real-time edge performance. Across domains, prevalent threats include DDoS, unauthorized access, and malware; persistent challenges include device heterogeneity, rapid exploit weaponization, nonstandardized evaluation, concept drift, adversarial/poisoning risks, and governance and privacy constraints that hinder real world rollouts. We conclude that ML materially strengthens IoT resilience but requires rigorous, industry-scale validation, lightweight and explainable models, protocol-aware designs, robust federated aggregation, and SDN/NFV orchestration; we outline benchmark and deployment priorities to translate laboratory gains into operational security.

Algorithms doi: 10.3390/a19050383

Authors: Yifei Deng Yudong Zhang Jingpu Yang Miao Fang

Maintaining high-fidelity character personas and tracking trusted narrative facts remain significant challenges for LLM-based role-playing systems, particularly in long-context scenarios. Traditional Retrieval-Augmented Generation (RAG) approaches, which typically rely on static, stateless retrieval, often struggle to capture evolving plot dynamics, leading to character hallucinations and logical inconsistencies over prolonged interactions. To address these limitations, we present FictionRAG, a novel stateful retrieval-augmented framework designed to enhance long-narrative role-playing. FictionRAG introduces a hierarchical memory architecture that decouples narrative information into three distinct lanes: factual events, persona traits, and worldview constraints. Furthermore, it employs a failure-driven metacognitive regulatory loop that dynamically identifies and corrects retrieval deficiencies—such as persona drift or conflicting world rules—before response generation. By treating role-playing as a dynamic state tracking problem rather than simple question answering, FictionRAG ensures that generated responses are strictly grounded in both the narrative timeline and the character’s psychological profile. Extensive experiments on a dataset comprising twenty classic novels demonstrate that FictionRAG significantly outperforms existing baselines in factual accuracy, persona stability, and worldview consistency. Beyond literary role-playing, these results suggest that stateful, evidence-constrained retrieval can serve as a general mechanism for long-form controllable generation tasks that require persistent state tracking and multi-dimensional consistency.

Algorithms doi: 10.3390/a19050382

Authors: Shanshan Peng Dandan Wang Fang Zhu

The demand for the fabric warehouse presents obvious characteristics of hockey stick effect. This leads to problems such as peak congestion and labor shortages during its operation. In order to alleviate this phenomenon, we propose a combination strategy that uses a SARIMA–Markov hybrid model for demand forecasting, and then applies Actor-Critic reinforcement learning for dynamic pricing. This model integrates SARIMA with Markov chains for residual correction, capturing linear trends and seasonal patterns while correcting residuals, yielding more accurate predictions for highly volatile demand in textile logistics. Experimental results indicate that our approach achieves better performance than SARIMA, Temporal Fusion Transformer (TFT), and Ensemble, especially in identifying and reproducing sharp demand peaks. By combining forecasting results with price elasticity, the proposed dynamic pricing scheme cuts peak-hour demand by 12.54%, which in turn eases pressure on labor scheduling and boosts the efficiency of workforce allocation. This work offers a data-driven approach to flattening demand fluctuations via intelligent pricing, improves operational efficiency without requiring extra hardware investment, and provides a practical response to a long-standing bottleneck in the textile logistics sector.

Algorithms doi: 10.3390/a19050380

Authors: Yuchen Huo Longyun Zhang Huijuan Guo Jingyi Gong Liqun Kuang Xie Han Fengguang Xiong

High-quality correspondences are critical to the accuracy and robustness of point cloud registration. Existing Transformer-based methods are fundamentally constrained by the quadratic computational complexity of self-attention, resulting in limited scalability. Moreover, conventional outlier removal paradigms operate by pruning initial correspondences, and thus fail catastrophically in low-overlap scenarios where initial inliers are inherently scarce. To address these challenges, we propose MaLCA, a point cloud registration method based on Mamba-enhanced features and local correspondence augmentation. We first adopt KPFCN as the backbone to extract multi-scale geometric features from raw point clouds. A Mamba selective state space model then replaces self-attention for global context modeling with linear complexity, while cross-attention is retained to facilitate inter-point-cloud feature interaction. Rather than following the conventional subtraction-based outlier removal paradigm, we introduce a prior-guided local rematching strategy combined with a fused neighbor matching mechanism that iteratively constructs dense, high-quality correspondences from sparse initial inliers, fundamentally overcoming the bottleneck of inlier scarcity in challenging scenes. Extensive experiments on the 3DMatch/3DLoMatch and 4DMatch/4DLoMatch benchmarks demonstrate that MaLCA achieves competitive registration performance across both rigid and deformable scenarios, with particular advantages in low-overlap cases.

Algorithms doi: 10.3390/a19050381

Authors: Abdullah Al-Refai Mohammad F. Al-Hammouri Bandi Vamsi Ali Al Bataineh

The prediction accuracy of RNA’s tertiary structure remains a core challenge in the field of computational biology. Existing models frequently encounter significant challenges due to the complexities of diverse topologies and the intricate nature of long-range interactions. We introduce RNAFoldDiff, a generative framework that integrates a sequence-aware graph transformer with a geometric diffusion process for end-to-end RNA 3D structure prediction. RNA sequences and secondary structures are converted into graph representations that capture backbone connectivity and base pair topology. The transformer models local motifs and global dependencies, while the diffusion module iteratively denoises coordinates into physically consistent conformations. The model was pretrained on more than 15,000 structural motifs from the RNA 3D Hub and fine-tuned on complete RNAs from the RNA-Puzzles dataset. In benchmarking tests, RNAFold-Diff achieved an average root mean square deviation (RMSD) of 2.64 Å, a Global Distance Test (GDT) score of 68.7%, and a base pair accuracy of 89.5%, reducing RMSD by nearly 30% and improving GDT by 9 points compared to RoseTTAFoldNA. The framework also outperformed FARFAR2, SimRNA, and RNAformer. Ablation experiments confirmed the contributions of diffusion refinement, edge-aware graph encoding, and motif-level pretraining, while qualitative analyses showed biologically plausible folds including helices, junctions, and multiloops. By combining topology-aware graph learning with generative diffusion, RNAFoldDiff advances RNA tertiary structure modeling and provides a practical tool for RNA design, ribozyme analysis, and structure-guided drug discovery.

Algorithms doi: 10.3390/a19050379

Authors: Jimmy Aurelio Rosales Huamani Jose Ogosi Auqui Pedro Toribio Pando Ernan Capcha Milla Jorge Luis Quinto Esquivel Jose Luis Castillo Sequera

Quantum computing is one of the most promising emerging technologies, and quantum machine learning (QML), as one of its key branches, is attracting growing interest for intelligent data processing in education. This study conducted a systematic review of QML in the context of Education 5.0 using the PRISMA 2020 methodology. A total of 48 peer-reviewed articles from Springer, Scopus, IEEE Xplore, PubMed, MDPI, arXiv, and APS were analyzed. The results indicate that QML has significant potential to enhance personalized learning, optimize educational data processing, support curriculum innovation, and foster the development of quantum-related competencies. Representative QML algorithms, including Quantum Support Vector Machines, variational quantum circuits, and quantum neural networks, are identified as key technological enablers for future educational applications. However, significant challenges remain, such as limited access to quantum infrastructure, lack of specialized curricula, hardware constraints, and the need for interdisciplinary training. Overall, this study highlights the growing relevance of QML for adaptive learning, learning analytics, and intelligent educational systems, while emphasizing the need for further empirical validation and scalable implementation in real educational environments.

Algorithms doi: 10.3390/a19050378

Authors: Zhijun Shen Qian Guo Lizhou Jiang Jingkang Huang Zhenfan Yu Xinlei Cai Hailin Pang Tao Yu

Power documentation serves as the core guideline for the safe operation of power systems, and its precise retrieval is crucial for ensuring grid stability and safety. In this context, Retrieval-Augmented Generation (RAG) frameworks emerge as an effective technique by combining LLMs with natural language understanding capabilities and a retrieval-based model with traceability. However, existing Retrieval-Augmented Generation (RAG) frameworks face several main challenges for power-system documents: semantic drift caused by non-standardized industry terminology, increased semantic noise due to fixed-window segmentation, and knowledge conflicts in the multi-source retrieval context. To address these challenges, we propose a multi-path adaptive fusion retrieval framework based on small language models (SLMs). To map queries to standard terminology, our framework first constructs a common terminology repository and section-structure-aware index for the power industry while fully preserving the physical hierarchical logic from related documents. Subsequently, the SLM in our framework assigns prior weights based on query features and retrieved context, which contributes to adaptive fusion of retrieval paths through confidence assessment and consistency verification. With the help of the fusion process, our method effectively filters retrieval noise and resolves knowledge conflicts. Experimental results on real-world power-document datasets covering dispatch, energy storage and emergency response show that our framework achieves an average recall of 91%, outperforming DENSE and BM25 by 21% and 28% respectively. Compared with other methods, it yields the optimal BERTScore F1 (0.7798) and Rouge-1/2/L F1 (0.2430, 0.1588, 0.2098) and achieves the best results in the RAGAS framework evaluation, which significantly enhances the rigor and reliability of the question-answering system in the power engineering domain.

Algorithms doi: 10.3390/a19050377

Authors: Zi Cheng Mengting Yuan Lefei Zhang

Equality saturation explores equivalent program expressions via e-graphs, and its extraction step selects one representative per equivalence class to form an output tree. Standard extraction minimizes a decomposable per-node cost function that cannot capture multi-node structural patterns arising in SMT preprocessing and compiler instruction selection. We formalize pattern-aware extraction as a weighted pattern cover problem on AND-OR DAGs and establish its correspondence to tree covering in instruction selection. Three challenges arise when migrating tree covering to e-graphs: annotation ambiguity from multiple candidates per class, context-dependent selection from depth-2 templates, and DAG sharing conflict. We show that the coupled selection–tiling problem reduces to a tree DP with three mutually exclusive tile-role states, generalizing BURS tree covering from fixed trees to AND-OR DAGs. A bottom-up pass computes optimal DP values, and a top-down pass traces back decisions to produce the output tree. For template depth at most two, the algorithm computes an exact optimum in O(N·K·|P|·Cmax) time. The evaluation targets extraction-level coverage, since end-to-end performance additionally depends on rewrite-rule design and saturation completeness. On SMT-COMP benchmarks, the algorithm achieves up to 31× higher weighted pattern coverage than standard extraction. Depth-2 tiling contributes 45–51% additional improvement, with overhead within 1.5× of standard extraction.

Algorithms doi: 10.3390/a19050376

Authors: Hao Hu Yujie Liu Xiaojin Jin Bo Hu

Accurate prediction of the Remaining Useful Life (RUL) of special gas cylinders is critical for industrial safety management. However, the nonlinear, strongly coupled degradation behaviors of these cylinders, combined with non-stationary and high-noise monitoring data, limit the performance of single deep learning models. Traditional hyperparameter tuning and signal processing methods often fail to meet the required prediction accuracy. To address these challenges, this study proposes a hybrid SSA–PSO–ResNet–LSTM–Attention framework for RUL prediction of special gas cylinders. The framework first applies Singular Spectrum Analysis (SSA) to decompose and reconstruct the 12-dimensional multi-source sensor signals, effectively suppressing noise while extracting core degradation trends. Subsequently, a ResNet–LSTM–Attention collaborative model is constructed, where ResNet ensures stable spatial feature propagation, LSTM captures long- and short-term temporal dependencies, and a multi-head attention mechanism emphasizes critical time steps associated with abrupt degradation. Furthermore, a Particle Swarm Optimization (PSO) algorithm is employed to globally optimize key hyperparameters, including the number of convolutional kernels, LSTM hidden units, and learning rate, mitigating the subjectivity of manual tuning. Experimental validation is conducted on 1000 real monitoring samples from 100 composite material gas cylinders, with a cylinder ID-based 7:1:2 train–validation–test split and stratified sampling covering four operating conditions. PSO optimizes hyperparameters using the validation set RMSE as the fitness function, and the test set is exclusively used for final performance evaluation. All results are reported as the mean ± standard deviation from grouped 5-fold cross-validation on the cylinder-wise partition. The proposed model achieves a test RMSE of 71.55, MAE of 50.63, and R2 of 0.9584, representing a 34.2% and 30.2% reduction in RMSE and MAE, respectively, compared with the second-best CNN-LSTM model, and significantly outperforming SVR, MLP, and other benchmark models. Ablation studies confirm the positive synergistic effect of each component, with the removal of either the attention mechanism or the ResNet module causing substantial performance degradation. By employing physically calibrated RUL labels and a balanced multi-condition dataset, the proposed framework achieves high predictive accuracy and good potential for industrial application, providing an effective solution for RUL prediction of special gas cylinders and similar high-pressure vessels, with potential applications in intelligent maintenance of complex industrial equipment.

Algorithms doi: 10.3390/a19050375

Authors: Paul A. Gagniuc

This work presents a compact arithmetic formulation for inverting row-major linear indices into three-dimensional coordinates. The formulation defines a bijective and reversible mapping based solely on integer division and modulo operations and avoids iteration and control-flow constructs. A traversal-based reconstruction strategy and the arithmetic formulation are evaluated on Graphics Processing Unit (GPU) hardware across multiple volumetric configurations. The experimental results show that arithmetic index decomposition yields uniform execution behavior, low run-to-run timing variability, and constant per-thread execution cost under massively parallel execution. The observed differences follow from GPU architectural characteristics, particularly sensitivity to control-flow divergence. The formulation provides a portable reference model for multidimensional index inversion suitable for parallel kernels and hardware-oriented implementations.

Algorithms doi: 10.3390/a19050374

Authors: Huijuan Liao Yanlong Sun

Alternation preference in random-sequence judgments has been linked to objective differences in pattern waiting times. The present study asks whether recurrent networks can learn such temporal asymmetries beyond single-pattern regularities and capture the more complex competitive structure of Penney’s game. To address this question, we adopt Penney’s game as a mathematically tractable testbed, in which competitive advantage is determined not by marginal sequence frequency but by the joint effect of self-overlap and cross-overlap structure. Based on Conway’s formula, we formulate two complementary tasks for gated recurrent units (GRUs): optimal counterstrategy prediction and win-probability estimation. Experimental results show that the GRU achieves strong performance on both tasks, recovering optimal or near-optimal second player responses and accurately estimating theoretical winning probabilities with good ranking consistency. These findings suggest that recurrent networks can learn structural regularities underlying asymmetric sequence competition, extending from single-pattern waiting-time effects to more complex competitive sequence settings.

Algorithms doi: 10.3390/a19050373

Authors: Tengda Guo Junjiang Zhu Yunjie Li

T-wave inversion (TWI) on an electrocardiogram (ECG) is a key indicator of myocardial ischemia, yet existing inverse ECG methods lack quantitative physiological parameter resolution. This study aims to propose a novel multiscale computational framework to inversely identify the ionic mechanisms underlying TWI. A cell–tissue–torso cardiac electrophysiological model was integrated with a differential evolution (DE) algorithm. The forward model combined the Grandi atrial model and BPS2020 ventricular model, simulating action potential propagation via cellular automata and body surface ECGs via field point potentials. The inverse solution optimized 29 physiological parameters by minimizing the root-mean-square error between the simulated and clinical ECGs. The method was applied to 30 normal and 30 TWI cases to analyze the repolarization abnormalities. The study revealed that extracellular Ca2+ > 2.88 mmol/L and K+ < 3.4 mmol/L in ventricular myocytes (Endo, M, Epi) induce TWI. Quantitative analysis identified specific 95% confidence intervals for ionic imbalances in three scenarios: Case 1 (VEpi>VEndo>VM) with [Ca2+] 2.60–3.30 mmol/L and [K+] 1.9–4.7 mmol/L; Case 2 (VM>VEpi>VEndo) with [Ca2+] 2.36–3.68 mmol/L and [K+] 3.13–4.07 mmol/L; and Case 3 (VEpi>VM>VEndo) with [Ca2+] 2.67–3.91 mmol/L and [K+] 3.11–3.45 mmol/L. This approach enables cellular-scale mechanistic insights into TWI by quantifying ionic concentration changes. The framework supports the advancement of personalized cardiac diagnostics and drug development.

Algorithms doi: 10.3390/a19050372

Authors: Manuel R. Arahal Manuel G. Satué Francisco Colodro Alfredo P. Vega-Leal

Switching losses make up for a notable portion of all losses in converter-supplied electric drives. Control algorithms such as Finite State Model Predictive Control (FSMPC) have tackled this issue in different ways; in particular incorporating a switching penalty to the cost function. This, however, results in an optimization problem with increased computational load, restricting the attainable sampling frequency for a given computing hardware. Recently, fast algorithms have been developed that reduce the computational load. However they cannot incorporate the switching penalty term. This paper explores a way around this problem for the particular case of stator current control of a five-phase induction motor. The proposal achieves fast computation even if a term for switching frequency reduction is present in the cost function. Experimental results show how stator current tracking performance is affected in both the torque producing plane and the harmonic subspace.

Algorithms doi: 10.3390/a19050371

Authors: Liliya A. Demidova Vladimir E. Zhuravlev

Accurate extrinsic calibration of multi-camera systems is a central problem in three-dimensional computer vision, as errors in the relative positioning of sensors directly propagate into geometric distortions that critically degrade the quality of downstream applications. This paper proposes an incremental extrinsic camera parameter initialization method that improves upon the baseline iterative registration algorithm based on the Perspective-n-Point (PnP) problem. Unlike board-based calibration frameworks, the proposed approach operates on individually placed markers with no prior knowledge of their mutual positions, enabling recalibration without dedicated calibration sessions. The accuracy improvement is achieved through the introduction of heuristic weighting of fiducial marker detections using AprilTags, as well as the application of a multi-view triangulation algorithm for dynamic refinement of marker spatial coordinates at each stage of scene expansion. Theoretical analysis demonstrates that the incorporation of these mechanisms does not increase the overall asymptotic computational complexity of the complete calibration cycle (including the global optimization stage), despite the higher computational cost of the initialization stage itself. Empirical validation of the method is performed on both synthetic datasets with known ground-truth camera parameters and real-world capture data through the evaluation of geometric errors and their comparison with the baseline method. Experimental results, supplemented by an ablation study, indicate that the proposed algorithm achieves statistically significant improvements on synthetic data in more than 80% of cases, while on real data it is on average 85% more accurate in terms of reprojection error.

Algorithms doi: 10.3390/a19050370

Authors: Paolo Marcandelli Stefano Mariani Martina Siena Stefano Markidis

Fourier representations play a central role in operator learning for partial differential equations and are increasingly being explored in quantum machine learning architectures. The classical fast Fourier transform (FFT), particularly in its Cooley–Tukey decomposition, exhibits a structure that naturally matches continuous-variable quantum circuits. This correspondence establishes a direct structural isomorphism between the Cooley–Tukey butterfly network and Gaussian photonic gates, enabling the FFT to be realized as a native optical computation in continuous-variable quantum computing. Building on this observation, we introduce a continuous-variable Quantum Fourier Layer (CV–QFL) based on a bipartite Gaussian encoding and a Cooley–Tukey quantum Fourier transform, enabling exact two-dimensional spectral processing within a Gaussian photonic circuit. We test the CV–QFL on two representative tasks: spectral low-pass filtering and Fourier-domain integration of the heat equation. In both cases, the results match the classical reference to machine precision. More broadly, this work lays the foundation for continuous-variable approaches to quantum scientific computing and for the development of native spectral architectures in quantum machine learning.

Algorithms doi: 10.3390/a19050369

Authors: Mikhail Kolev Irina Naskinova Mariyan Milev Hristo Kalinov Gabriela Vasileva Penko Mitev

The relationship between blood glucose levels and blood pressure is well established in clinical literature, yet its precise quantitative characterization, including nonlinear effects, threshold phenomena, and demographic modifiers, remains incompletely understood. In this study, we conducted a comprehensive cross-sectional analysis of the National Health and Nutrition Examination Survey (NHANES) spanning 11 survey cycles (1999–2023), comprising 54,881 adult participants with at least one glycemic marker and standardized blood pressure measurements. Of these, 26,981 had valid fasting plasma glucose (FPG) measurements, and 49,327 had valid glycated hemoglobin (HbA1c) measurements. We employed restricted cubic splines (RCS), generalized additive models (GAMs), and segmented regression to characterize the dose–response relationship between glycemic markers and both systolic (SBP) and diastolic blood pressure (DBP). A 10 mg/dL increase in FPG was associated with a 0.32 mmHg increase in SBP (95% CI: 0.26–0.38, p < 0.001) after adjusting for age, sex, and body mass index (BMI). Nonlinearity was statistically significant for all exposure–outcome combinations (p < 10−7 for Wald tests). Segmented regression identified a FPG breakpoint at 122.1 mg/dL (95% CI: 119.5–125.6), below which SBP increased at 0.39 mmHg per mg/dL and above which the association was essentially flat. Stratified analyses revealed that the glucose–BP association was strongest in females (β = 0.048 per mg/dL) compared with males (β = 0.021), and in prediabetic individuals (β = 0.065) compared with those with established diabetes (β = 0.014). In the statistical mediation decomposition, body mass index accounted for 23.5% of the total FPG–SBP association. A significant FPG × BMI interaction (p < 0.001) indicated that the glucose–BP relationship is modulated by adiposity. These findings provide a large-scale population-level analysis of the glucose–blood pressure dose–response relationship and identify potential thresholds warranting further investigation for integrated cardiometabolic risk management (95% bootstrap CI: 19.3–28.9%; 1000 resamples); given the cross-sectional design and BMI’s plausible role as a shared upstream determinant of glucose and blood pressure, this proportion is reported as a confounding decomposition rather than as evidence of causal mediation. Insulin resistance (HOMA-IR) and C-reactive protein did not contribute significantly as additional decomposition pathways.

Algorithms doi: 10.3390/a19050368

Authors: Audrey Rah Yuhua Chen

Early-stage stroke rehabilitation increasingly incorporates virtual reality (VR) systems to provide interactive motor training and positive reinforcement. However, the minimal voluntary plantar pressure activations generated during early recovery are often below the detection limits of conventional pressure-sensing platforms, restricting timely feedback. This study quantitatively evaluates the detectability of low-amplitude plantar micro-intent signals under varying sensor resolution and adaptive threshold conditions. Publicly available plantar pressure recordings from the PhysioNet Center for Verification and Evaluation of Stroke (CVES) database were used as physiological baseline signals. Micro-intent was modeled as short-duration half-sine pressure pulses with systematically varied amplitudes and integrated into low-load baseline segments. Sensor resolution was represented through controlled noise modeling to emulate low-, medium-, and high-resolution sensing scenarios. A sliding-window adaptive threshold detector was evaluated across multiple amplitudes and sensitivity stages. The detection probability, false positive rate, and minimum detectable amplitude (defined as ≥80% detection probability) were quantified. The results show that detection probability increases with signal amplitude and shifts toward lower amplitudes with improved sensor resolution and more sensitive threshold configurations. Higher-resolution sensing reduced the minimum detectable amplitude, while adaptive thresholding enabled earlier detection of weak plantar activations without substantial increases in false positives. These findings provide quantitative design guidance for pressure-sensing VR rehabilitation systems targeting early-stage motor recovery.

Algorithms doi: 10.3390/a19050367

Authors: Zhadyra Alimbayeva Chingiz Alimbayev Kassymbek Ozhikenov Kairat Karibayev Aiman Ozhikenova Ussen Shylmyrza Dilfuza Akhmedova

The aim of this study was to evaluate the feasibility of predicting percutaneous coronary intervention (PCI) based on clinical, laboratory, and electrocardiographic data available at various stages of hospitalization. A retrospective single-center study was conducted, including 137 patients with suspected coronary artery disease. The fact that PCI was performed during the current hospitalization was considered as the endpoint. Taking into account the temporary availability of data, three sets of signs were formed: basic (SAFE), including indicators available at admission; clinical (CLINICAL); and extended (EXTENDED), supplemented with glycemic parameters. Logistic regression, random forest, and gradient boosting were used to build the models. The assessment was carried out using repeated stratified cross-validation (5 × 10). The main metrics were ROC-AUC, PR-AUC, accuracy and F1-measure. The models demonstrated moderate predictive ability. The basic model (SAFE) showed a ROC-AUC of 0.734 ± 0.092, while the best results were achieved using an extended model based on a random forest (ROC-AUC 0.755 ± 0.079). The addition of glycemic parameters provided a moderate improvement in prediction quality. In the logistic regression, the most significant predictor was the presence of type 2 diabetes mellitus (OR = 7.36; p < 0.001). The results indicate the potential for using non-invasive data to assess the likelihood of PCI in the early stages of hospitalization. However, the models show moderate accuracy and require further validation on larger and more independent samples.

Algorithms doi: 10.3390/a19050366

Authors: Enes Rayman Serra Öğütcen Okan Yaman Yusuf Murat Erten

Natural disasters are devastating occurrences that have a major influence on the well-being of numerous individuals on a global scale. The primary goal of this study is to facilitate the rapid, transparent, and safe delivery of various aid such as food and clothing to people in disaster areas. For this purpose, a system has been established using blockchain technology in cooperation with institutions and humanitarian organizations. This system is designed to be accountable and reliable; it will supervise all processes from the source of aid materials to their distribution while protecting the personal information of disaster victims. The assistance process is improved using Smart Contracts in order to provide fast, effective, and coordinated assistance. Unlike existing humanitarian frameworks that rely on permissionless networks such as Bitcoin or Ethereum, this study proposes Hyperledger Fabric to ensure beneficiary privacy and eliminate per-transaction fees for end-users, thereby offering a more sustainable economic model for high-frequency aid distribution compared to public blockchains. The proposed system (Chainguard) addresses the ’efficiency gap’ in the current literature JSON Web Token (JWT)-based authentication layer. The results showed that Chainguard achieves a stable throughput of ~180 TPS with an end-to-end latency of less than 1.5 s, outperforming traditional heavy-cryptography models in terms of scalability and resource efficiency during real-time disaster response.

Algorithms doi: 10.3390/a19050365

Authors: Vincent Roberge Mohammed Tarbouchi

This paper presents a parallel genetic algorithm (GA) for the planning of power distribution networks considering harmonics. Power distribution systems are generally operated in a radial configuration, supplemented by tie switches that enable network reconfiguration during unexpected outages or planned maintenance. They can also include distributed generators (DGs), capacitor banks (CBs), and soft open points (SOPs) to lower distribution losses and improve the voltage profile. Some of the loads and DG units may be nonlinear, generating harmonic currents in the system, polluting the power, and increasing losses. This paper makes use of a parallel GA to find an optimized configuration, optimized location, and sizing of DGs, CBs, and SOPs to lower real power distribution losses while considering harmonics and the physical constraints of the network. The proposed algorithm uses a solution encoding based on the minimum spanning tree to guarantee the radial topology of candidate solutions. It uses the backward–forward power flow method to compute the fundamental voltages and a decoupled harmonic power flow for the harmonic components. The algorithm is parallelized on a small computer cluster using the Message Passing Interface (MPI) to reduce its execution time. The proposed solver is validated on distribution systems ranging from 16 to 880 buses. The results show that simultaneously optimizing the topology, the DGs, the CBs, and the SOPs results in reducing power losses by 37% to 93%, improving the overall efficiency of the distribution system. The parallelization using MPI allows for a 90.9× speedup on a 96-core cluster.

Algorithms doi: 10.3390/a19050364

Authors: Fangyuan Jin Hui Lin Lu Zhang Yiwei Chen

Vision–language models (VLMs) are increasingly deployed in resource-constrained environments, yet efficient fine-tuning remains challenging because post-training quantization often degrades the effectiveness of low-rank adaptation. This paper revisits that mismatch in the context of MobileVLM1.7B and presents QuantFT-VL, a novel initialization strategy following the quantization phase to seamlessly align with the LoRA technique. The key idea is to initialize LoRA using a low-rank approximation of the quantization residual instead of the default zero-initialization used in QLoRA-style pipelines. After quantizing a pretrained weight matrix W into Q, we compute the residual W − Q and use truncated singular value decomposition to initialize the LoRA factors (A and B) so that the starting adapted weight Q + ABT better matches the full-precision model. This residual-aware initialization reduces the discrepancy introduced by quantization and leads to faster and more stable optimization. Experiments on six standard VLM benchmarks show that QuantFT-VL consistently improves over QLoRA and recovers performance close to or better than full-precision LoRA in the best setting. On two RTX 3090 GPUs, QuantFT-VL improves the average benchmark score by 3.27 percentage points over QLoRA while preserving the memory and speed advantages of quantized fine-tuning.

Algorithms doi: 10.3390/a19050363

Authors: Oscar H. Sierra-Herrera Mario Eduardo González Niño Edwin Francis Cárdenas Correa Jersson X. Leon-Medina Francesc Pozo

While hardware selection is widely recognized as a key factor in machine learning performance, systematic and reproducible evaluation across heterogeneous and accessible platforms remains limited, particularly when jointly considering execution time, energy consumption, stability, and cost-efficiency. This work presents a unified and fully reproducible benchmarking framework for supervised learning, designed to enable controlled and comparable evaluation across diverse hardware environments. The proposed methodology enforces consistent training pipelines, fixed hyperparameter configurations, and repeated executions to ensure statistical reliability, while incorporating performance metrics such as execution time, power consumption, and energy usage, as well as performance-per-dollar. The framework is validated on a representative set of platforms, including CUDA-enabled GPUs, Apple Silicon (CPU/GPU), x86 processors, ARM-based embedded systems, and cloud-based environments, using convolutional, recurrent (RNN, LSTM, BiLSTM), and tree-based (XGBoost) models. The results reveal that hardware efficiency is strongly model-dependent. GPUs provide the highest computational performance and stability for parallel workloads, whereas Apple Silicon achieves superior energy efficiency with competitive execution times, particularly for recurrent architectures. The batch size analysis shows that performance can vary significantly depending on workload configuration, especially on CPU-based platforms, while epoch-based evaluation confirms that the measured performance reflects steady-state behavior rather than initialization overhead. In contrast, conventional CPUs and embedded systems exhibit significant scalability limitations for deep learning training, although they remain competitive for tree-based methods such as XGBoost, which demonstrates near hardware-independent predictive performance. These findings highlight the limitations of generalized hardware selection criteria and emphasize the need for model-aware and hardware-aware benchmarking. The proposed framework offers a practical and extensible foundation for reproducible, hardware-aware evaluation of machine learning systems, supporting informed decision-making in research, deployment, and cost-constrained scenarios.

Algorithms doi: 10.3390/a19050361

Authors: Muhammad Amir Saeed Antonio Candelieri

Bayesian optimization (BO) is a widely used framework for optimizing expensive black-box functions. Most BO methods rely on Gaussian process (GP) surrogates, which perform well in continuous domains but encounter difficulties when decision variables include categorical or mixed discrete–continuous components. In particular, GP-based approaches typically require ad hoc numerical encodings of categorical variables that may fail to capture the structure of discrete decision spaces. In this work, we propose MNL-BO (Multinomial Logit Bayesian Optimization), a preference-based Bayesian optimization framework that replaces the GP surrogate with a multinomial logit (MNL) model trained from pairwise preference comparisons. The resulting surrogate provides a natural and interpretable representation of categorical alternatives while allowing continuous, discrete, and categorical variables to be handled within a unified optimization framework. The predictive utility estimates and uncertainty indicators generated by the MNL model are employed to formulate acquisition functions that reconcile exploration with exploitation. The proposed methodology is evaluated on three progressively complex optimization challenges: a purely categorical benchmark, a combinatorial Traveling Salesman problem, and a constrained mixed-variable engineering design problem concerning material selection in pressure vessel optimization. Multi-run tests provide consistent advantages over random search and exhibit stable convergence behavior across diverse random initializations. In addition to heuristic baselines such as local search and classical metaheuristics, we also compare against tree-based Bayesian optimization baselines inspired by the Sequential Model-based Algorithm Configuration (SMAC) framework. The results indicate that the proposed MNL-BO method achieves competitive performance under comparable evaluation budgets while providing an interpretable probabilistic surrogate for categorical decision spaces. These findings suggest that preference-based surrogate modeling provides a practical and flexible alternative for Bayesian optimization in categorical and mixed-variable optimization problems.

Algorithms doi: 10.3390/a19050360

Authors: Ramiro Esquivel-Felix Mireya Moreno-Lucio Celina Lizeth Castañeda-Miranda Héctor Alonso Guerrero-Osuna Rodrigo Castañeda-Miranda Carlos A. Olvera-Olvera Ma. del Rosario Martínez-Blanco Luis Octavio Solís-Sánchez

In precision agriculture, identifying pests such as the whitefly (Bemisia tabaci) is a significant challenge, as precise knowledge of these insects is essential for developing effective Integrated Pest Management (IPM) strategies. Automated daily monitoring within IPM programs optimizes the diagnostic registration stage by reducing logistical expenses and manual errors, enabling early pest treatment interventions and providing quantitative data for informed decision-making. In this study, an image bank was processed using a Quaternionic Gabor Filter (QGF) algorithmto highlight textural features through hypercomplex correlation. The highlighted objects were then processed by a YOLOv8 pretrained model to identify Bemisia tabaci. Experimental results demonstrate that this combination achieves a precision of 0.868 and an mAP@0.5 of 0.950, while a PSNR of 34.10 dB ensures the structural integrity of the enhanced images. Although the total execution time averages 2.3 s per image due to preprocessing complexity, the GPU inference time of 10.3 ms confirms the potential for high-speed detection. This approach significantly enhanced the morphological features of Bemisia tabaci, increasing the robustness of the detection model and narrowing down processing conditions for yellow trap samples to strengthen precision in the semi-arid regions of Zacatecas, Mexico.

Algorithms doi: 10.3390/a19050362

Authors: Chouvalit Khancome

Multiple string matching is a fundamental operation in real-time analytics, cybersecurity, bioinformatics, and large-scale information retrieval. Nevertheless, existing approaches continue to face inherent trade-offs among preprocessing efficiency, verification overhead, and support for dynamic pattern updates, particularly in large and continuously evolving environments. This paper presents MMIVL, a high-performance algorithm founded on the multi-character inverted list (m-CIVL), a unified and inherently dynamic indexing framework for pattern management. By integrating positional information, termination semantics, and pattern associations within a single structure, m-CIVL enables direct matching without requiring a separate verification stage. MMIVL achieves a preprocessing complexity of O(|P|/s), a search complexity of O(|T| + nocc), and an update complexity of O(|p|/s), where s denotes the segment length. Extensive experiments on synthetic and real-world datasets demonstrate that MMIVL consistently outperforms representative baselines, with especially strong gains in large-scale scenarios, while maintaining stable performance and favorable memory efficiency. Overall, these results establish m-CIVL as an effective, scalable, and practically viable solution that unifies efficient preprocessing, high-throughput searching, and dynamic update capability for modern multiple string-matching applications.

Algorithms doi: 10.3390/a19050359

Authors: Luca Bergamaschi Michele Bergamaschi

We derive the eigenvalue bounds for symmetric block-tridiagonal multiple saddle-point systems preconditioned with the symmetric positive definite (SPD) preconditioner proposed by J. Pearson and A. Potschka in 2024 and further studied by L. Bergamaschi and coauthors, and for double saddle-point problems with inexact Schur complement matrices. The analysis applies to an arbitrary number of blocks. We validate the proposed estimates with both synthetic and realistic test problems, and show the good performance of the proposed preconditioner under the condition that the Schur complements are accurately approximated.

Algorithms doi: 10.3390/a19050358

Authors: Roberto Stevens Porto Solano Antonio Berlanga de Jesús José M. Molina Yair Rivera Julio

While most current work on interpretable models has centered on post hoc explainability of individual predictive models, the structure of the hypothesis space from which such models are drawn has been largely neglected. This paper proposes a contrasting perspective in which explainability is treated not as an attribute of a single solution but as a structural property of the model space. By combining Rashomon set analysis with Pareto-based performance–model complexity trade-offs, we formulate a computational framework for identifying near-optimal and structurally simple models. A performance–model complexity trade-off landscape is constructed by systematically generating models under controlled complexity bounds and extracting Pareto-optimal solutions. The results show that explainability can emerge as a regional property of hypothesis spaces in which multiple interpretable models achieve competitive predictive performance. This perspective supports the identification of robust and auditable predictive solutions and complements traditional explainability approaches centered on isolated models. Cross-dataset replication on Wine (UCI) and Vehicle (UCI) confirms the generalizability of these findings.

Algorithms doi: 10.3390/a19050357

Authors: Fengkai Ye Ruoqian Li Danping Wang Mengyang Li

Accurate forecasting of agricultural product prices is crucial for informed decision-making in agricultural markets; however, such time series are inherently characterized by non-stationarity, multi-scale dynamics, and substantial noise, posing significant challenges to conventional methods. To overcome these limitations, this study proposes a novel hybrid framework, termed TOC-CNN-BiLSTM-SA, built upon a “quadratic decomposition–clustering–optimization” paradigm. Specifically, a composite CEEMDAN–K-means++–VMD approach is first employed to hierarchically decompose the raw price series via coarse decomposition, feature clustering, and refined decomposition, enabling effective noise suppression and multi-scale feature extraction. Subsequently, a deep learning architecture integrating Convolutional Neural Networks (CNNs), Bidirectional Long Short-Term Memory networks (BiLSTM), and a self-attention mechanism is developed, where CNN captures local patterns, BiLSTM models bidirectional temporal dependencies, and the attention mechanism enhances global feature representation. Furthermore, the Tornado Optimizer with Coriolis force (TOC) is introduced to adaptively tune key hyperparameters, thereby improving model robustness and generalization capability. Empirical results based on wheat price data from Henan Province, China, demonstrate that the proposed model achieves outstanding predictive performance, with RMSE, MAE, MAPE, and R2 values of 4.425, 3.9372, 0.16%, and 99.97%, respectively, significantly outperforming existing benchmark models. These research indicate that the proposed framework effectively captures complex price dynamics and offers a reliable and practical solution for agricultural price forecasting.

Algorithms doi: 10.3390/a19050356

Authors: Francisco Javier Moreno Arboleda Luzarait Cañas Quintero Georgia Garani

Blockchain systems generate massive volumes of transactional data, yet most existing analytical approaches rely on query-based retrieval mechanisms that treat transactions as isolated records. In this paper, a trajectory-based framework for blockchain analysis is introduced where user activity is modeled as temporally ordered behavioral patterns. Four types of blockchain trajectories are formally defined: miner reward trajectories, sender value-and-fee trajectories, receiver value trajectories, and sender–receiver interaction trajectories. Unlike traditional query frameworks, trajectories are treated as first-class analytical objects, explicitly constructed and returned as outputs, thereby enabling structured temporal reasoning over blockchain behavior. To demonstrate the practicality of the approach, the proposed trajectory functions are implemented in Python 3.12 and experiments are conducted using real data from the Ethereum blockchain. Compared with conventional query-based approaches that return isolated transactions, the experimental results show that the proposed trajectory-based framework enables a more systematic identification of temporal behavioral patterns, including persistent miner dominance, recurrent zero-value interactions, sender–receiver role reversals and sender dominance by sending the highest values across several periods. The results show that trajectory-based modeling provides a systematic lens for uncovering temporal and structural regularities that are not readily observable through conventional query techniques. This work establishes a formal foundation for behavioral blockchain analytics and opens new research directions in centralization measurement, predictive modeling, and trajectory similarity analysis.

Algorithms doi: 10.3390/a19050355

Authors: Tianyi Wang Tao Shen Zhiyuan Zhang Shuangping Huang Huiguo He Qingguang Chen Houqiang Yang

Accident reports provide a detailed account of environmental causes, unsafe human behaviors, and subsequent chain reactions. These records serve as essential resources for analyzing accident mechanisms and exploring potential risk patterns within production safety processes. Currently, Graph based Retrieval-Augmented Generation (RAG), which integrates Large Language Models (LLMs) with Knowledge Graphs (KGs), has emerged as a leading approach for complex causal question answering over extensive unstructured accident documentation. However, the application of this technology in the production safety domain still encounters two primary challenges. First, knowledge graph construction using a single granularity fails to capture fine-grained case details and macro-level standard systems. Second, traditional one-step retrieval paradigms lack the capacity to track deep causal chains or interpret the complex logic of multi-factor coupling. To address these limitations, we propose CausalAgent, a hierarchical graph-enhanced multi-agent framework for causal question answering in production safety accident reports. This framework innovatively combines a Hierarchical Causal Graph (HC-Graph) and a Multi-Agent Collaborative Reasoning (MACR) mechanism. Specifically, the HC-Graph employs a two-layer architecture that links a fine-grained instance layer with a national standard causation layer to resolve conflicts in semantic granularity. The MACR mechanism converts complex natural language queries into executable structured queries and logic verification steps through the sequential cooperation of four specialized agents, namely the Graph Parsing Agent, the Problem Analysis Agent, the Query Generation Agent, and the Reasoning Insight Agent. CausalAgent enables in-depth mining of accident causation mechanisms and provides scientific, robust and interpretable intelligent support for data-driven risk assessment and emergency decision-making. Experiments on real-world accident datasets demonstrate that CausalAgent achieves a 100.0% query execution rate and an 87.3% reasoning accuracy, outperforming the SOTA baseline by 45.2% in terms of absolute accuracy.

Algorithms doi: 10.3390/a19050354

Authors: Spiros Thivaios Georgios Kostopoulos Antonia Stefani Sotiris Kotsiantis

Financial fraud has become a critical challenge for modern financial systems due to the rapid growth of digital transactions, online banking services, and electronic payment platforms. Traditional rule-based fraud detection systems are increasingly inadequate in addressing the evolving and adaptive strategies employed by fraudsters. Consequently, Machine Learning (ML) and Deep Learning (DL) techniques have emerged as powerful tools for detecting fraudulent activities in large-scale financial datasets. This paper presents a comprehensive survey of ML/DL approaches for financial fraud detection. The survey systematically reviews existing research across multiple methodological paradigms, including classical supervised learning, anomaly detection, graph-based methods, deep neural networks, multimodal architectures, and cost-sensitive learning frameworks. Particular emphasis is placed on emerging techniques such as graph neural networks, transformer-based architectures, and federated learning approaches designed to address privacy and scalability challenges. In addition to reviewing model architectures, this work analyzes key challenges inherent to fraud detection systems, including extreme class imbalance, concept drift, adversarial behavior, data privacy constraints, and real-time deployment requirements. Furthermore, the survey examines evaluation methodologies, highlighting the limitations of commonly used metrics and discussing more realistic evaluation strategies that incorporate operational costs and risk management considerations. This paper also provides a structured taxonomy of fraud detection methods, comparative analyses of commonly used datasets, and a synthesis of current research trends. Finally, open challenges and promising research directions are identified, including adaptive learning systems, interpretable Artificial Intelligence models, graph-based behavioral modeling, and privacy-preserving collaborative fraud detection frameworks.

Algorithms doi: 10.3390/a19050352

Authors: Isaac Oliva-González Hugo Jiménez-Hernández

The shortest path problem presents formidable challenges in graph optimization, particularly within dense or large-scale networks, where traditional algorithms face serious scalability limitations. This paper puts forth a robust QUBO-based simulated annealing (QUBO-SA) methodology that effectively utilizes a Quadratic Unconstrained Binary Optimization (QUBO) framework to encode path costs and structural constraints simultaneously. Our approach has been rigorously evaluated on synthetic graphs with controlled connectivity, varying from n=10 to n=40, and on a real-world urban transportation network from Querétaro, Mexico, comprising n=443 nodes. We assess performance through rigorous probabilistic reliability indicators, notably the success probability psuccess, Time-to-Solution, and the relative runtime ratio R(ptarget), benchmarked against Dijkstra’s algorithm. In small synthetic instances (n=10), the QUBO-SA method demonstrates outstanding success rates (psuccess≥0.97) with runtimes on par with the deterministic baseline (R0.99≈1). However, as the problem size increases, success probabilities diminish while computational overhead rises, with R0.99 soaring from approximately 1.0 at n=10 to between 4.63 and 5.83 at n=40. For the urban network, our solver achieves success probabilities between 0.49 and 0.91, depending on the specified path length, with R0.99 values ranging from 2.17 to 9.41. Notably, reducing the target confidence level from 99% to 90% cuts runtime overhead by approximately fifty percent across all configurations. Although the QUBO formulation demonstrates scalability in relation to n+m, potentially limiting its use in dense graphs, the sparse structure typical of real-world road networks enables competitive performance in moderately large instances. These findings decisively highlight the trade-off between solution reliability and computational efficiency, pinpointing specific problem regimes where QUBO-based optimization methods are not only viable but advantageous for path-optimization tasks.

Algorithms doi: 10.3390/a19050353

Authors: Mostafa Zahed Maryam Skafyan Morteza Rasoulianboroujeni

The application of machine learning (ML) models to accelerate the discovery of high-transfection-potency ionizable lipids has gained significant momentum in advancing lipid nanoparticle (LNP)-mediated RNA delivery. In the present study, we adopt a screening-oriented evaluation framework based on early-recognition ranking metrics tailored to high-throughput discovery. Model performance was assessed using the enrichment factor (EF), normalized discounted cumulative gain (NDCG), and HitRate at the top 10% of the ranked list, with uncertainty quantified via 1000 nonparametric bootstrap resamples. To assess robustness of conclusions, additional analyses were conducted at the top 1% and top 5% thresholds, reflecting increasingly stringent prioritization scenarios. Four predictive models—XGBoost, Random Forest, Elastic Net, and Quantile Regression Forest—were evaluated across three molecular feature representations, circular Morgan fingerprints, expert-crafted descriptors, and Grover graph embeddings, using a held-out test set. Across all models and thresholds, Morgan fingerprints consistently yielded superior early-recognition performance. The best-performing configuration—XGBoost with Morgan fingerprints—achieved EF@10% = 4.850 (95% CI [3.182, 6.818]), NDCG@10% = 0.628 (95% CI [0.234, 0.909]), and HitRate@10% = 0.493 (95% CI [0.318, 0.683]), corresponding to nearly fivefold enrichment over random selection and identification of highly potent lipids in approximately half of the prioritized candidates. Threshold-sensitivity analyses revealed that although stricter cutoffs (top 1% and top 5%) exhibit greater variability, the relative performance ordering of molecular representations remains stable. Bootstrap distributional comparisons further demonstrate that Morgan fingerprints provide not only higher but also more consistent screening performance than expert descriptors and Grover embeddings. Collectively, these results indicate that molecular representation—rather than model architecture—is the primary determinant of early-recognition performance in ionizable lipid discovery and that this conclusion is robust across multiple screening depths.

Algorithms doi: 10.3390/a19050351

Authors: Carlos Gabriel Mireles-Preciado Diana Carolina Toledo-Pérez Roberto Augusto Gómez-Loenzo Marcos Aviles Juvenal Rodríguez-Reséndiz

Objective: This study investigates the optimization of surface EMG (sEMG) classification for seven transtibial movements using short analysis windows (64 ms) suitable for real-time control of below-knee prostheses. Methods: We systematically evaluated feature engineering strategies, dimensionality reduction techniques, and classification approaches using linear Support Vector Machines on four-channel sEMG data from the transtibial region. We compared amplitude-based versus derivative-based time-domain features, integrated frequency-domain features, and implemented multi-window majority voting with 50% overlap. Results: Evaluated across nine subjects (four male, five female), the optimized system achieves a population-level accuracy of 70.16%±7.09% with multi-window majority voting (per-subject range: 60.71–78.57%), with voting consistently improving accuracy over single-window classification by +7.06% on average. We demonstrate that PCA provides zero benefit for linear classifiers when all features are retained. Documented failed approaches include adaptive windowing and spectral entropy features. Conclusion: Careful feature engineering combining time-domain (MAV2, RMS, VAR, MAX, LOG, IEMG) and frequency-domain features (MPF, MF, band powers) with multi-window voting substantially recovers accuracy losses from aggressive window reduction while maintaining sub-100 ms latency suitable for prosthetic control. This work provides a validated methodology across multiple subjects for optimizing EMG classification latency–accuracy trade-offs, demonstrates that PCA is unnecessary for linear classifiers with well-engineered features, and documents negative results to guide future prosthetic control research.

Algorithms doi: 10.3390/a19050350

Authors: Epameinondas Panagopoulos Charalampos M. Liapis Anthi Adamopoulou Ioannis Kamarianos Sotiris Kotsiantis

This study investigates how teacher-education undergraduates in Greece use, evaluate, and trust Artificial Intelligence (AI) in higher education, with particular attention to the gap between widespread adoption and limited epistemic trust. The topic is important because generative AI is rapidly entering universities, reshaping learning practices, academic integrity, and the legitimacy of knowledge, while learners often rely on systems whose outputs are not easily verifiable. The study focuses on future teachers because they are both current users of AI in higher education and likely future mediators of its use in school settings. Addressing this problem, the study contributes empirical evidence on how AI adoption relates to epistemic authority and institutional legitimacy within teacher education rather than across university students in general. A mixed-methods design was employed using a structured questionnaire completed by 363 teacher-education undergraduates from the University of Patras and the University of Ioannina in Greece; the sample was predominantly women (86.0%) and first-year students (92.6%). Quantitative responses were analyzed statistically, open-ended answers were examined thematically, and factor analysis was used to identify latent attitudinal dimensions. The findings indicate very high AI use in everyday life (92.6%) and study practices (81.3%), but only moderate trust: 1.4% reported complete trust and 12.1% generally trusted AI-generated answers. Six dimensions explained 61.73% of total variance, pointing to a layered attitudinal structure within this teacher-education population, consistent with an adoption–trust paradox and with the need for transparent, verifiable, human-supervised educational AI. The observed verification-based trust calibration may partly reflect an emerging pedagogical orientation toward source checking and responsibility for knowledge mediation, but given the strong concentration of first-year students, this should be interpreted as characteristic of early-stage teacher education rather than of university students more broadly.

Algorithms doi: 10.3390/a19050349

Authors: Chuanping Bai Donglin Jing Zhixue Wang Fangqin Zhang

The scale features of rock art exhibit significant diversity and graduality. Among the existing semantic segmentation methods for rock art, although some models have taken note of the scale differences in rock art patterns and the complexity of directional features, and proposed targeted improvement strategies, most of these methods view scale adaptation and directional representation as unconnected problems. They fail to model the intrinsic correlation between the scale adaptation and directional representation, and particularly overlook the restrictive effect of scale accuracy on the extraction of directional features. This ultimately leads to the problem of “spatial representation misalignment” in the semantic segmentation of rock art. To address the above problems, this paper proposes a Dynamic Fine-tuning Rotation Network (DFTR-Net), which aims to solve the problems of imprecise scale feature extraction and directional misalignment for rock art patterns with arbitrary orientations. The network consists of a dynamic selective convolution structure and a shapeaware spatial feature extraction module. Specifically, the dynamic selective convolution dynamically adjusts the coverage range of the receptive field through inter-layer feature aggregation. It uses stacked small dilated convolution kernels to replace large convolution kernels with the same receptive field for extracting the neighborhood details of patterns. Then, by combining with feature aggregation, it constructs spatial feature differences and realizes intra-layer dynamic weighted fusion, thereby achieving accurate scale feature extraction. After obtaining fine-grained scale features, the shape-aware module first corrects the initial segmentation candidate regions of the patterns to generate directional guide boxes. Subsequently, it drives the rotational sampling of convolution kernels based on the angles of the guide boxes, forming region-constrained deformable convolutions that adapt to the shape of the patterns. These convolution kernels obtain strong supervision based on pixel-level annotations, which enhances the sensitivity to the directional features of the patterns and effectively alleviates the problem of directional misalignment. Extensive experiments show that DFTR-Net can achieve higher performance on the 3D-pitoti and Petroglyph Annotation datasets compared with the existing methods.

Algorithms doi: 10.3390/a19050347

Authors: Shruti Kulkarni Alexios Mylonas Stilianos Vidalis

Threat modelling is a methodology employed for identifying and analysing threats and applicable mitigations for web applications, mobile applications, infrastructure, and environments including smart home environments. Threat modelling starts with a tabletop exercise to identify threats. It provides extremely important insights into what can go wrong if certain events or a series of events take place. The identification of these events is critical to ensuring the right mitigation strategies are applied. Threat modelling also helps to identify security controls that may be assumed to provide required security, but, in reality, may not be addressing the existing and applicable threat(s). Existing literature, in the public domain and in academia, discusses threat materialisation for smart homes; however, entry points for a threat to materialise and exploit these vulnerabilities are not explored and a dedicated threat model for smart home environments is currently unavailable. Whilst threats can be mitigated by smart home device manufacturers, there are also mitigations that need to be applied by smart home owners who are both technology-aware and technology-unaware. In this paper, we propose a structured, domain-specific threat modelling methodology for smart home environments. The methodology models threats from a smart home owner’s perspective, identifies entry points and the mitigations that need to be implemented by a smart home owner. It also acknowledges that the attack surface expands and contracts and is not constant; which is addressed by applying zero-trust principles.

Algorithms doi: 10.3390/a19050348

Authors: Zihao Wang Li Huang Hua Han Mingyang Chen

Large-scale constrained multi-objective optimization problems (LCMOPs) pose significant challenges due to the curse of dimensionality, complex constraint landscapes, and high computational overhead. In time-sensitive scenarios, existing large-scale constrained multi-objective evolutionary algorithms (LCMOEAs) often incur high computational costs and therefore struggle to meet efficiency requirements. This paper proposes a GPU-accelerated tensorized flexible differential evolution algorithm (TFDEMO) for LCMOPs. To address the curse of dimensionality and complex constraint landscapes in LCMOPs while maintaining GPU-level parallel efficiency, a tensorized flexible differential evolution operator (FlexDE) is developed. It utilizes a Bernoulli masking mechanism to switch between guided and random mutation modes in parallel on the GPU. The guidance probability is adaptively adjusted based on historical performance and the evolutionary state. Furthermore, a dual-population collaborative neighborhood selection mechanism is designed. For the main population, a Boolean mask tensor method is proposed, which constructs four Boolean mask tensors in parallel to encode feasibility states and dominance relations across all subproblems and their neighborhoods, and aggregates them via bitwise operations to produce the dominance tensor in a single pass. The auxiliary population performs constraint-ignoring neighborhood selection and shares its offspring with the main population to assist the main population in crossing large infeasible regions. The experimental results on the LIRCMOP and ZXH_CF benchmark suites with decision variable dimensions ranging from 100 to 800 demonstrate that TFDEMO achieves the best overall performance among the compared algorithms under both fixed-time and fixed function-evaluation settings. Additionally, a portfolio rebalancing problem with three objectives, five constraints, and scalable dimensions is designed to evaluate the performance of the proposed algorithm in time-sensitive application scenarios.

Algorithms doi: 10.3390/a19050346

Authors: Nusirat Ojuolape Gold Husain Coovadia

Automated long-essay scoring (ALES) is gradually considered as a means to enhance efficiency and consistency in large-scale assessment; however, concerns remain regarding its suitability, particularly as it relates to the reliability, validity, and fairness of ALES-assigned grades relative to human-grades in high-stakes professional contexts. This study examines these concerns using over 15,000 long essay examination scripts from a professional accounting certification examination. The study examines whether the ALES confidence index (CI) meaningfully predicts grading accuracy or points to systemic grading failures. Findings reveal fair overall agreement between human and ALES grades, with high within ±1 grade agreement, and rare yet task-concentrated ALES grading failures, while CI shows statistically significant but practically weak predictive value and limited discrimination. The results support the use of ALES as an assistive, human oversight tool rather than an independent grader, highlighting the importance of task-based validation, stronger calibration analysis, and continuous human supervision in high-stakes professional assessment contexts. The study advances innovative assessment practices, but calls for cautious deployment of ALES and recommends integration of a hybrid human-in-the-loop approach, multi-disciplinary validation, and capacity building to strengthen ethical and responsible AI usage in accounting education and professional practice, aligning with SDGs 4 and 9.

Algorithms doi: 10.3390/a19050345

Authors: Paul A. Gagniuc

Antivirus systems have evolved from static pattern matchers into complex algorithmic ecosystems that encapsulate the broader logic of modern cybersecurity. This review deconstructs their internal architecture, tracing the transition from deterministic string-matching automata to probabilistic, behavioral, and cloud-assisted paradigms. Foundational modules such as scanners, heuristic analyzers, behavioral monitors, and sandbox environments operate as interconnected computational strata, forming adaptive feedback loops that mirror principles of distributed intelligence. Signature-based methods, such as Aho-Corasick, Boyer-Moore, and Wu-Manber, remain core to real-time filtering, while probabilistic reasoning through Bayesian inference, Markov modeling, and Hidden Markov Models extends detection to polymorphic and metamorphic threats. Behavioral analysis, empowered by Support Vector Machines, deep neural architectures, and temporal models, enables semantic inference over system-call graphs and runtime telemetry. Moreover, cloud-assisted frameworks integrate federated learning and global reputation graphs, which transform detection into a collective intelligence process.

Algorithms doi: 10.3390/a19050344

Authors: Dongping Tian Jie Sun Fang Li Yuyu Fan Xiaorui Gou Siyu Peng Zhongzhi Shi

Particle swarm optimization (PSO) is an optimizing method that is based on the theory of swarm intelligence. PSO is an effective algorithm that is used to search in a parallel manner compared to other methods. However, PSO has a tendency towards local optima when tackling complex multimodal optimization problems. It also has the disadvantages of slow convergence process and poor stability in the latter evolutionary period. In view of these demerits, a hybrid PSO method based on chaotic opposition-based initialization and an adaptive learning strategy is presented in this work (abbreviated as ACMPSO). First, the chaos initialization and opposition-based learning (OBL) are employed to produce high-quality initial particles in the feasible region, which is able to improve the quality of the initial solutions. Second, the logistic mapping embedded inertia weight is formulated to better trade off the global and local search process. Third, the global optimal particle is regulated by an exclusive velocity and position updating strategy whereas the rest particles are adjusted by the standard updating mechanism so as to prevent particles from premature convergence. Furthermore, an adaptive position update paradigm is developed to finely regulate the global exploration and local exploitation. Finally, conducted experiments on CEC’13 and CEC’22 reveal that the proposed ACMPSO outperforms several other advanced PSO variants regarding their convergence rate and accuracy. Alternatively, to further illustrate the effect of ACMPSO, we have applied it to two real-world problems, and simulation results ascertain its effectiveness and robustness.

Algorithms doi: 10.3390/a19050343

Authors: Shuo Sun Haiyang Yu

Corner detection is a fundamental task in computer vision that plays a critical role in applications such as image registration, 3D reconstruction, and object tracking. In biological visual systems, simple cells in the primary visual cortex exhibit high selectivity to edge stimuli of specific orientations, while end-stopped cells can detect geometric singular structures such as line segment endpoints and corners. Existing corner detection methods based on visual neural computation typically employ a strategy of densely distributed end-stopped cells for corner localization, which suffers from significant localization deviation under small angle conditions due to mutual interference between responses of adjacent neurons. To address this problem, this paper proposes a bionic corner detection method based on cooperative processing of simple cells and end-stopped cells. The method constructs a two-stage cooperative processing framework: the edge filtering stage employs a Gabor filter bank to simulate the orientation selectivity of simple cells, extracting edge positions and orientation information; the dynamic construction stage builds unilateral end-stopped cells only at filtered edge positions based on local orientation information, fundamentally avoiding computational redundancy and response interference caused by global dense distribution; the corner localization stage determines precise corner coordinates through hierarchical clustering and dual-cluster centroid fusion strategies. Experimental results demonstrate that, in the 15° acute-angle regime where dense end-stopped schemes are most severely affected by response interference, the proposed method reduces the mean localization error from 8.76 to 2.34 pixels, corresponding to a 73.3% improvement; averaged across the eight tested angle levels from 15° to 165°, the improvement is approximately 40.9%, and all per-angle differences are statistically significant (paired t-test, p < 0.01 or below, N = 10 independent runs). On standard test images, the method attains the lowest mean localization error among the eight compared detectors (1.58 pixels, versus 1.68–3.42 pixels for Harris, FAST, COSFIRE, KAZE, SuperPoint, Deep Corner, and Wei et al.), while maintaining competitive detection rate, false-alarm rate, and runtime. Physiological plausibility validation experiments show that the correlation coefficient between the detection deviation of this method and human perceptual deviation reaches 0.923, indicating that the output of the framework aligns with previously reported human perceptual bias patterns and supporting its biological plausibility as a biologically inspired—rather than mechanistic—model of corner perception. The source code, dataset, and experimental results are publicly available (see Data Availability Statement).

Algorithms doi: 10.3390/a19050342

Authors: Oswaldo González-Gaxiola Jehan Saleh Ahmed Lina S. Calucag Anjan Biswas

This paper utilizes the Laplace–Adomian decomposition method to numerically investigate the highly dispersive bright soliton solutions in twin-core optical couplers that employ metamaterials as waveguides. The focus of the study is on the power-law self-phase modulation. The results of the simulations and the accompanying error analysis demonstrate exceptional accuracy for this numerical approach. These findings suggest that the Laplace–Adomian decomposition method is a robust tool for tackling complex nonlinear problems in optical systems. Furthermore, the implications of this research could pave the way for advancements in the design and optimization of metamaterial-based waveguides, potentially leading to improved performance in applications, such as telecommunications and sensing technologies.

Algorithms doi: 10.3390/a19050341

Authors: Fangyuan Jin Hui Lin Lu Zhang Yiwei Chen

Single-image dehazing is still a challenging problem because haze mainly corrupts low-frequency structures such as global contrast and color consistency, while fine textures and object boundaries are degraded in a different manner. In this paper, we present a frequency-guided multi-scale dehazing network (FGDNet) that explicitly couples spatial-domain restoration and Fourier-domain feature decomposition in a compact U-Net-like architecture. Built on a gated U-Net backbone, the proposed model inserts a frequency processing branch into encoder stages. In detail, the feature maps are transformed by fast Fourier transform, split into low- and high-frequency components through a radial mask, refined separately, and fused by a lightweight cross-domain gating module. The low-frequency pathway emphasizes color and illumination recovery, whereas the high-frequency pathway enhances edges and textures. Moreover, an additional Fourier amplitude supervision term aligns the spectral distribution of restored images with haze-free targets. Experimental results on RESIDE ITS, RESIDE OTS, O-HAZE, and NH-HAZE show that the proposed method achieves 33.3 dB PSNR/0.983 SSIM on ITS, 35.1 dB PSNR/0.988 SSIM on OTS, 19.1 dB PSNR/0.786 SSIM for OTS-trained generalization to O-HAZE, and 15.8 dB PSNR/0.648 SSIM for OTS-trained generalization to NH-HAZE. Furthermore, both quantitative and qualitative results demonstrate that the proposed method provides a more effective and more robust solution than representative dehazing methods. In addition, ablation studies confirm that both the Fourier branch and the spatial–spectral gating mechanism contribute consistently to performance gains. These results support the effectiveness of explicit frequency-aware representation learning for image dehazing and suggest a practical direction for improving generalization from synthetic to real haze.

Algorithms doi: 10.3390/a19050340

Authors: Mostafa Aboulnour Salem Zeyad Aly Khalil

This study proposes a scalable Explainable Artificial Intelligence (XAI)–driven Bayesian–AI decision–control framework for multi-objective optimisation in uncertain and dynamic systems. The framework integrates Bayesian networks, stochastic control, and expected utility theory within a unified probabilistic architecture. Unlike traditional black-box models, the proposed framework provides intrinsic interpretability through probabilistic reasoning and dependency-aware modelling. This allows users to understand how decisions are formed and how variables influence outcomes. To further strengthen explainability, the framework incorporates post hoc XAI techniques, including SHAP-based feature attribution and sensitivity-based local explanations. These methods quantify the contribution of each variable and provide clear explanations at both global and local levels. The system is formulated as a stochastic state-space model and implemented as a closed-loop adaptive architecture. It updates decisions continuously as new data becomes available. Scalable inference is achieved using variational inference, Markov Chain Monte Carlo, and Sequential Monte Carlo methods. This ensures efficient performance in complex and high-dimensional environments. A simulation study based on 370 observations shows that the proposed framework improves decision quality, robustness under uncertainty, and transparency compared to conventional methods. Explainability is evaluated using Fidelity, Stability, and Transparency metrics. The results confirm that the model produces consistent and reliable explanations. The framework supports human-centred decision-making by providing visual analytics and clear probabilistic explanations. This makes it suitable for high-stakes applications such as cyber–physical systems, intelligent platforms, and real-time AI systems. The main contribution of this study is the integration of intrinsic probabilistic interpretability with post hoc XAI techniques into a single, scalable framework. This approach bridges a key gap in XAI research and offers a practical and transparent solution for decision-making under uncertainty.

Algorithms doi: 10.3390/a19050339

Authors: Zaijiang Yu He Jiang Yan Zhao

In the current energy context, enhancing the precision of wind power prediction serves as a key enabler for the stable development of the power grid. In the existing wind power prediction models, there are often problems of modal aliasing and noise residue, or the prediction accuracy of the model is not high. In an effort to solve the problem of short-term wind power forecasting, a wind power series decomposition and reconstruction method based on improved complete ensemble empirical mode decomposition with adaptive noise-variational modal decomposition (ICEEMDAN-VMD) secondary decomposition is proposed. Using ICEEMDAN, wind power data (wind direction, wind speed, temperature, humidity, air pressure, etc.) is decomposed into several IMF sub-series, and these IMF sub-series are categorized into three different frequency components by combining sample entropy, Q statistics and sequence frequency. Secondly, the gray wolf optimization (GWO) is improved by using the empirical exchange strategy (EES), and the optimization performance of the EES-GWO proposed in this paper is verified by using 10 test functions. Finally, the EES-GWO-convolutional neural network–bidirectional gated recurrent unit–global attention (EES-GWO-CNN-BiGRU–Global attention) high-frequency component prediction model is constructed. Finally, we employ the XGBoost model to forecast the mid- and low-frequency components, thereby generating the corresponding forecasting results. The support vector machine (SVM) model nonlinearly integrates all the forecasting results to produce the final forecasting results. Through example analysis and comparison, the performance of the proposed model is verified from two perspectives.

Algorithms doi: 10.3390/a19050338

Authors: Suziee Sukarti Mohamad Fani Sulaima Norashikin Sahadan Muhamad Hafizul Shamsor Siaw Wei Yao Aida Fazliana Abdul Kadir

Reliable quantification of industrial energy savings requires accurate detection of non-routine events (NREs) that distort post-retrofit baselines. Conventional statistical and rule-based anomaly detection methods often misinterpret operational variability, leading to biased or overstated savings under the International Performance Measurement and Verification Protocol (IPMVP). This study develops a novel IPMVP-compliant hybrid deep learning framework that integrates a deterministic Deep Neural Network (DNN) for baseline modeling with stochastic architectures, namely the Factored Conditional Restricted Boltzmann Machine (FCRBM) and Generative Adversarial Network (GAN), to capture probabilistic reconstruction patterns. Their outputs are fused using a hybrid thresholding mechanism that balances detection sensitivity and specificity. Using high-resolution data from an industrial glove manufacturing facility, the hybrid DNN–FCRBM model achieved the best trade-off, demonstrating an accuracy of 94.3%, a precision of 91.1%, and a low false positive rate of 5.1%. This model validated 11.32% industrial energy savings (approximately 478,050 kWh), equivalent to 237 tonnes of CO2 avoided. The integration of stochastic generative learning within a deterministic framework strengthens transparency, auditability, and IPMVP compliance, offering a scalable pathway for credible industrial energy savings verification.

Algorithms doi: 10.3390/a19050337

Authors: Tong Wu Meihao Zhu Zhansheng Liu Xiaofeng Zhang Fengjuan Chen Xiaoqing Zhu Haowen Sun Chuan Zhang Jiahao Wu

Path planning for inspection and maintenance robots in nuclear power plants often suffers from limited adaptability, high computational cost, and unstable convergence in obstacle-dense confined environments. To address these issues, this paper proposes an improved Bi-RRT–APF path optimization framework for complex industrial scenarios. The method integrates (1) a hybrid sampling strategy combining random, goal-biased, and potential-field-guided sampling to enhance global exploration and convergence efficiency; (2) a potential-field-guided perturbation and stagnation detection mechanism to improve escape capability from local minima; and (3) a dynamic target switching and constrained segmented connection strategy to improve path feasibility and safety. A digital twin-based simulation platform is further developed to validate the engineering applicability of the proposed approach. Simulation results demonstrate significant quantitative improvements over baseline methods. Compared with conventional RRT and Bi-RRT, the proposed method reduces iteration count by 65.3% and 43.8%, respectively, and decreases computation time by 76.1% and 48.4%, respectively, while increasing the success rate to 95% (from 82% and 93%) and improving path smoothness (reduced from 5.3 and 3.3 to 2.9). Compared with advanced variants (Quad-RRT and KB-RRT*), the method further reduces computation time by 25.2% and 10.3% and iteration count by 29.3% and 8.4%, respectively. These results indicate that the proposed method achieves a balanced improvement in efficiency, robustness, and path quality. This work provides an efficient and reliable solution for autonomous path planning of robots in complex nuclear power plant environments.

Algorithms doi: 10.3390/a19050336

Authors: Li Hao Ma Ning

Imaging-based multi-omics derived from digital histopathology provides a valuable approach for characterizing tumor heterogeneity from routine clinical specimens. However, robust multi-cancer histopathological analysis remains challenging due to pronounced intra-tumor variability, inter-organ morphological overlap, and sensitivity to staining and acquisition variations, which can limit the generalizability of deep learning models. These limitations are largely driven by insufficient representation learning, particularly in multi-organ and multi-class diagnostic settings. In this study, we propose a hierarchically regularized representation learning framework for multi-cancer histopathological image analysis that models imaging-based features across multiple organs and diagnostic categories. The framework integrates complementary mechanisms to capture fine-grained cellular morphology, long-range tissue architecture, and organ-aware diagnostic semantics within a unified computational model. A hierarchical supervision strategy guides the network to reduce entanglement between organ-level structural characteristics and disease-specific diagnostic patterns in the learned representations. The method operates without pixel-level annotations or handcrafted morphological priors, supporting scalable experimental evaluation. We demonstrate the approach on balanced lung and colon cancer histopathology cohorts, achieving 96.5% accuracy on lung cancer classification and 96.8% accuracy on colon cancer classification. Ablation and robustness analyses further validate the contributions of hierarchical regularization and consistency learning. Overall, this work provides a demonstrated proof-of-concept framework for representation-centric imaging-based analysis in multi-organ histopathology under the evaluated dataset conditions.

Algorithms doi: 10.3390/a19050334

Authors: Gerardo Iovane Giovanni Iovane Antonio De Rosa

Most neural architectures model time as a one-dimensional real-valued variable, constraining temporal reasoning to sequential propagation along a single axis. We introduce Complex-Time Neural Networks (CTNN), a new class of architectures in which temporal coordinates are elements of the complex plane T=t+iτ∈C, where ReT preserves chronological ordering and ImT encodes an orthogonal experiential dimension. Within this geometry, ImT<0 defines a memory domain enabling retrospective retrieval, ImT=0 corresponds to present-moment computation, and ImT>0 defines an imagination domain for prospective projection. We prove the Expressive Separation Theorem (Theorem 1), establishing that, within the temporally coupled function class GTCP and under explicit Assumptions A1–A4 (in particular the bounded projection Assumption A3), CTNN accesses temporally coupled functions at O(1) cost with respect to temporal distance Δ1, Δ2, while real-time architectures incur Ω(Δ1 + Δ2) sequential steps. For layered compositions, this yields an exponential composition gap within GTCP under A1–A4. These advantages hold under the stated assumptions and may not directly generalize to broader function classes or large-scale settings where A3 cannot be maintained. Therefore, Theorem 1 provides a formal separation result for GTCP, while CTNN more broadly defines a geometric framework for temporal computation. As the first concrete instantiation of this framework, we develop Complex-Time Convolutional Neural Networks (CTCNN). CTCNN achieves state-of-the-art performance on Something-Something V2 (70.2±0.4%, +1.1% over VideoMAE v2, p<0.01), strong performance on Kinetics-400 (78.4±0.3%), and substantial gains on Long Range Arena Path-X (87.3% vs. 79.6%, +7.7%), using 3.4× fewer parameters than VideoMAE v2. Learnable angular parameters α and β provide computationally interpretable parameters related to memory-access span and prospection breadth, with values varying systematically across task families.

Algorithms doi: 10.3390/a19050335

Authors: Erik Martínez-Vera Pedro Bañuelos-Sánchez Alfredo Rosado-Muñoz Juan Manuel Ramirez-Cortes Pilar Gomez-Gil

Power converters are fundamental components in vehicle electrification systems. However, their inherently nonlinear and time-varying condition requires complex design procedures when conventional control strategies based on linear small-signal models are employed. This work proposes a simplified and hardware-oriented DC-DC converter control methodology that combines fuzzy logic and Neural Networks in a sequential manner. A fuzzy logic fuzzy controller is first used to generate a dataset of control actions under closed-loop operation. A lightweight neural network is then trained using the obtained data to approximate this mapping and subsequently replace the fuzzy controller in real-time operation. To validate the approach, a bidirectional buck–boost DC-DC converter is designed for applications in the powertrain of electric vehicles with 500 kHz switching frequency and 13 kW power rating. The control algorithm is embedded in an FPGA to demonstrate its suitability for hardware deployment. The experimental results show a reduction in RMSE of 33.7% and a decrease in the settling time of at least 51.7% when compared with a benchmark PID control.

Algorithms doi: 10.3390/a19050333

Authors: Cheng Chen Faisal N. Abu-Khzam Levente Dojcsak Michael A. Langston

A k-partite graph is one whose vertices can be partitioned into k disjoint partite sets, with edges allowed between but not within these sets. In such a graph, a maximal k-partite clique is a subgraph with at least one vertex from each partite set and every allowable edge such that the subgraph cannot be enlarged by the incorporation of additional vertices. A maximum k-partite clique is of course a maximal k-partite clique of the greatest size. The results reported here describe a novel and practical modification of the best previously published algorithm for the enumeration of these special subgraphs. The relative performance of this new method relies on implicit edge addition and search tree pruning and is evaluated on graphs constructed from both pseudorandom and real-world data.

Algorithms doi: 10.3390/a19050332

Authors: Nedelcho Ganchovski Oscar Smith Christopher Rackauckas Lachezar Tomov Alexander Traykov

The Modified Anderson–Björck method is a new, robust, and efficient bracketing root-finding algorithm. It combines bisection with the Anderson–Björk method to achieve both fast performance and worst-case optimality. It relies on linearity check criteria for switching methods and uses Anderson–Björk corrections to overcome the fixed endpoint issue of false-position. Initial benchmarks of this method have shown certain performance advantages compared to other methods, such as Ridders, Brent and ITP. In this paper, we propose further improvements to this method and perform some additional analysis and benchmarks of its behavior and performance.

Algorithms doi: 10.3390/a19050329

Authors: Fangya Tan Yang Zhou Shuqiao Li Chun Jiang Jian-Guo Zhou Srikar Bellur

(1) Background: Advances in machine learning (ML)-based survival modeling enable the analysis of high-dimensional biomedical data. However, many approaches rely on the proportional hazards (PH) assumption, which is frequently violated in oncology and can limit the interpretability of hazard ratio-based results. Using Estrogen Receptor (ER) status in the METABRIC breast cancer cohort as a case study, we propose a framework that integrates machine learning survival models with Restricted Mean Survival Time (RMST) to provide a more robust and clinically interpretable approach for survival analysis under non-proportional hazards. (2) Methods: Overall survival was analyzed in 1104 patients. PH violations were confirmed using Schoenfeld residuals and Kaplan–Meier inspection. We compared four models: stratified Cox Elastic Net (Cox E-Net), Random Survival Forest (RSF), Gradient Boosting Survival Analysis (GBSA), and DeepHit. Performance was assessed using Harrell’s C-index, time-dependent IPCW C-index, and Integrated Brier Score (IBS). RMST at 180 months was utilized to quantify absolute survival differences between ER subgroups. To improve the stability of the estimates, 200 bootstrap resamples were performed, and 95% confidence intervals were derived from the bootstrap distribution. (3) ER status demonstrated significant PH violation (p < 0.005) with crossing survival curves. Discrimination (C-index 0.664–0.725) and calibration (IBS 0.149–0.169) were comparable across models, with RSF achieving the highest overall performance. Despite similar accuracy, survival curve structures differed substantially. Cox E-Net and RSF reproduced the observed crossing pattern, whereas GBSA generated smoother trajectories and DeepHit showed marked compression of subgroup separation. In the independent test cohort, the empirical RMST difference at 180 months was 16.6 months (ER-positive: 130.4; ER-negative: 113.8). Model-based RMST differences ranged from 1 month (DeepHit) to 27 months (Cox E-Net), with RSF and GBSA (12.8 and 13.8 months) most closely approximating the empirical benchmark. (4) Conclusions: We propose a novel, model-agnostic ML + RMST framework that addresses non-proportional hazards while providing quantifiable, time-specific clinical benefit. Moreover, models with similar discrimination and calibration produced markedly different survival curve behavior and absolute RMST estimates, demonstrating that accuracy metrics alone are insufficient for clinical interpretation. By linking prognostic modeling with absolute survival quantification, this framework advances survival evaluation beyond relative risk ranking toward individualized, clinically meaningful decision support.

Algorithms doi: 10.3390/a19050331

Authors: Mustafa B. Nidham Khalid Yahya Mehdi Safaei Nawal Rai Saleh Al Dawsari

This article is an in-depth analysis of the performance and efficiency of various control systems used in quadrotor unmanned aerial vehicles (UAVs). The study is focused on the comparison of three main control approaches, including Sliding Mode Control (SMC), Fuzzy Logic Control (FLC), and an extended version of Sliding Mode Control with the use of the Gray Wolf Optimizer (SMC-GWO), as well as a supportive validation model the Genetic Algorithm (SMC-GA). Based on the Newton–Euler formulation, the mathematical model of a quadrotor has been developed to provide a true picture of the dynamic behavior of the quadrotor. The model was then implemented in MATLAB/Simulink 2025b to test the performance of the system in its nominal and perturbed conditions. The findings have shown that the hybrid SMC-GWO controller has significant improvement in response speed, accuracy, and stability compared to the other controllers. Precisely, the SMC-GWO demonstrated 78.46 percent decrease in rise time and 23.40 percent decrease in settling time compared to the traditional SMC, as well as a nearly negligible steady-state error (SSE = 0.0008) in the roll channel. The proposed controller in the pitch channel reduced the rise time by 93.65 percent and the settling time by 20.22 percent, with a much smoother and more stable tracking and an effectively negligible steady-state error (SSE = 0.0001). The hybrid controller in the yaw channel had a 77.94 percent better rise time and 23.16 percent better settling time, resulting in a steady-state error of 0.0022. In relation to altitude control, SMC-GWO decreased the rise time by 91.87 percent and settling time by 25.04 percent over classical SMC, yet the steady-state error was almost zero. Under constant, time-varying actuator disturbances, the SMC-GWO controller also demonstrated better system stabilization and trajectory-tracking behavior than both SMC and FLC, as well as slightly better behavior than SMC-GA in the presence of faults and disturbances. These results verify that a UAV control framework based on the combination of the Gray Wolf Optimizer and Sliding Mode Control is more resilient, quick, and significantly more precise.

Algorithms doi: 10.3390/a19050330

Authors: Milan Toma

The integration of artificial intelligence into medical diagnostics has evolved from controlled research demonstrations to real-world clinical deployment, creating both unprecedented opportunities and substantial challenges [...]

Algorithms doi: 10.3390/a19050328

Authors: Zhanel Baigarayeva Assiya Boltaboyeva Zhuldyz Kalpeyeva Raissa Uskenbayeva Maksat Turmakhan Adilet Kakharov Aizhan Anartayeva Aiman Moldagulova

Unlike reactive systems, mechanical ventilation controlled by CO2 concentration operates at a target efficiency that dynamically increases whenever the target CO2 level is exceeded. This approach eliminates the typical ‘dead-time’ and prevents air quality degradation by ensuring the system adjusts its performance immediately in response to concentration changes. In this work, the study focuses on the development and evaluation of data-driven predictive models for near-term indoor CO2 forecasting that can be integrated into pre-occupancy ventilation strategies, rather than designing a complete control scheme. Experimental data were collected over four months in a 48 m2 smart laboratory configured as an open-plan office, where a heterogeneous IoT sensing architecture logged synchronized time-series measurements of CO2 and microclimate variables (temperature, relative humidity, PM2.5, TVOCs), together with acoustic noise levels and appliance-level energy consumption used as indirect occupancy-related signals. Raw telemetry was transformed into a 22-feature state vector using a structured feature engineering method incorporating z-score standardization, cyclic time encodings, multi-horizon CO2 lags, rolling statistics, momentum features, and non-linear interactions to represent temporal autocorrelation and daily periodicity. The study benchmarks multiple regression paradigms, including simple baselines and ensemble methods, and found that an automated multi-level stacked ensemble achieved the highest predictive fidelity for short-term forecasting, with an Mean Absolute Error (MAE) of 32.97 ppm across an observed CO2 range of 403–2305 ppm, representing improvements of approximately 24% and 43% over Linear Regression and K-Nearest Neighbors (KNN), respectively. Temporal diagnostics showed strong phase alignment with observed CO2 rises during occupancy transitions and statistically reliable prediction intervals. Five-fold walk-forward cross-validation confirmed the temporal stability of these results, with top models achieving consistent R2 values of 0.93–0.95 across Folds 2–5. These results demonstrate that, within a single-room university laboratory setting, historical sensor data from low-cost IoT devices can support accurate short-term CO2 forecasting, providing a predictive layer that could support future proactive ventilation scheduling aimed at reducing CO2 lag at the start of occupancy while avoiding unnecessary ventilation runtime. Generalization to other building types and occupancy profiles requires further validation.