Search Results (16)

Accurate bearing fault diagnosis under various operational conditions presents significant challenges, mainly due to the limited availability of labeled data and the domain mismatches across different operating environments. In this study, an adaptive meta-learning framework (AdaMETA) is proposed, which combines dynamic task-aware model-independent meta-learning (DT-MAML) with efficient multi-scale attention (EMA) modules to enhance the model’s ability to generalize and improve diagnostic performance in small-sample bearing fault diagnosis across different load scenarios. Specifically, a hierarchical encoder equipped with C-EMA is introduced to effectively capture multi-scale fault features from vibration signals, greatly improving feature extraction under constrained data conditions. Furthermore, DT-MAML dynamically adjusts the inner-loop learning rate based on task complexity, promoting efficient adaptation to diverse tasks and mitigating domain bias. Comprehensive experimental evaluations on the CWRU bearing dataset, conducted under carefully designed cross-domain scenarios, demonstrate that AdaMETA achieves superior accuracy (up to 99.26%) and robustness compared to traditional meta-learning and classical diagnostic methods. Additional ablation studies and noise interference experiments further validate the substantial contribution of the EMA module and the dynamic learning rate components. Full article

This study proposes a hybrid EGARCH-Informer framework for forecasting volatility and calibrating tail risk in financial time series. The econometric layer (EGARCH) captures asymmetric and persistent volatility dynamics, while the attention layer (Informer) models long-range temporal dependence with sparse attention. The framework produces horizon-specific forecasts (H = 1 and H = 5) that are mapped to VaR and CVaR at α = 0.95 and 0.99. Evaluation covers pointwise accuracy (MAE, RMSE) and risk coverage calibration (CVaR bias and Kupiec’s unconditional coverage), complemented by Conditional Coverage (CC) and Dynamic Quantile (DQ) diagnostics, and distributional robustness via a Student-t mapping of VaR/CVaR. Across four U.S. equity indices (SPX, IXIC, DJI, SOX), the hybrid matches GARCH at the short horizon and yields systematic error gains at the longer horizon while maintaining higher calibration quality than deep learning baselines. MAE and RMSE values remain near 0.0002 at H = 1, with relative improvements of 2–6% at H = 5. CVaR bias stays tightly bounded; DQ rarely rejects, and CC is stricter but consistent with clustered exceedances, and the Student-t results keep the median hit rates near nominal with small, mildly conservative CVaR biases. These findings confirm the hybrid model’s robustness and transferability across market conditions. Full article

High-precision spiral trajectory tracking for aquaculture net-cage inspection is hindered by uncertain hydrodynamics, strong coupling, and time-varying disturbances acting on an underactuated autonomous underwater vehicle. This paper adapts and validates a model–data-driven learning-aided adaptive robust control strategy for the specific challenge of high-precision spiral trajectory tracking for aquaculture net-cage inspection. At the kinematic level, a serial iterative learning feedforward compensator is combined with a line-of-sight guidance law to form a feedforward-compensated guidance scheme that exploits task repeatability and reduces systematic tracking bias. At the dynamic level, an integrated adaptive robust controller employs projection-based, rate-limited recursive least-squares identification of hydrodynamic parameters, along with a composite feedback law that combines linear error feedback, a nonlinear robust term, and fast dynamic compensation to suppress lumped uncertainties arising from estimation error and external disturbances. A Lyapunov-based analysis establishes uniform ultimate boundedness of all closed-loop error signals. Simulations that emulate net-cage inspection show faster convergence, higher tracking accuracy, and stronger robustness than classical adaptive robust control and other baselines while maintaining bounded control effort. The results indicate a practical and effective route to improving the precision and reliability of autonomous net-cage inspection. Full article

Background: Myasthenia gravis (MG), a chronic autoimmune disorder with variable disease trajectories, presents considerable challenges for clinical stratification and acute care management. This systematic review evaluated machine learning models developed for prognostic assessment in patients with MG. Methods: Following PRISMA guidelines, we systematically searched PubMed, Embase, and Scopus for relevant articles published from January 2010 to May 2025. Studies using machine learning techniques to predict MG-related outcomes based on structured or semi-structured clinical variables were included. We extracted data on model targets, algorithmic strategies, input features, validation design, performance metrics, and interpretability methods. The risk of bias was assessed using the Prediction Model Risk of Bias Assessment Tool. Results: Eleven studies were included, targeting ICU admission (n = 2), myasthenic crisis (n = 1), treatment response (n = 2), prolonged mechanical ventilation (n = 1), hospitalization duration (n = 1), symptom subtype clustering (n = 1), and artificial intelligence (AI)-assisted examination scoring (n = 3). Commonly used algorithms included extreme gradient boosting, random forests, decision trees, multivariate adaptive regression splines, and logistic regression. Reported AUC values ranged from 0.765 to 0.944. Only two studies employed external validation using independent cohorts; others relied on internal cross-validation or repeated holdout. Of the seven prognostic modeling studies, four were rated as having high risk of bias, primarily due to participant selection, predictor handling, and analytical design issues. The remaining four studies focused on unsupervised symptom clustering or AI-assisted examination scoring without predictive modeling components. Conclusions: Despite promising performance metrics, constraints in generalizability, validation rigor, and measurement consistency limited their clinical application. Future research should prioritize prospective multicenter studies, dynamic data sharing strategies, standardized outcome definitions, and real-time clinical workflow integration to advance machine learning-based prognostic tools for MG and support improved patient care in acute settings. Full article

Background: Artificial intelligence (AI) and machine learning (ML) have been reshaping maternal, fetal, neonatal, and reproductive healthcare by enhancing risk prediction, diagnostic accuracy, and operational efficiency across the perinatal continuum. However, no comprehensive synthesis has yet been published. Objective: To conduct a scoping review of reviews of AI/ML applications spanning reproductive, prenatal, postpartum, neonatal, and early child-development care. Methods: We searched PubMed, Embase, the Cochrane Library, Web of Science, and Scopus through April 2025. Two reviewers independently screened records, extracted data, and assessed methodological quality using AMSTAR 2 for systematic reviews, ROBIS for bias assessment, SANRA for narrative reviews, and JBI guidance for scoping reviews. Results: Thirty-nine reviews met our inclusion criteria. In preconception and fertility treatment, convolutional neural network-based platforms can identify viable embryos and key sperm parameters with over 90 percent accuracy, and machine-learning models can personalize follicle-stimulating hormone regimens to boost mature oocyte yield while reducing overall medication use. Digital sexual-health chatbots have enhanced patient education, pre-exposure prophylaxis adherence, and safer sexual behaviors, although data-privacy safeguards and bias mitigation remain priorities. During pregnancy, advanced deep-learning models can segment fetal anatomy on ultrasound images with more than 90 percent overlap compared to expert annotations and can detect anomalies with sensitivity exceeding 93 percent. Predictive biometric tools can estimate gestational age within one week with accuracy and fetal weight within approximately 190 g. In the postpartum period, AI-driven decision-support systems and conversational agents can facilitate early screening for depression and can guide follow-up care. Wearable sensors enable remote monitoring of maternal blood pressure and heart rate to support timely clinical intervention. Within neonatal care, the Heart Rate Observation (HeRO) system has reduced mortality among very low-birth-weight infants by roughly 20 percent, and additional AI models can predict neonatal sepsis, retinopathy of prematurity, and necrotizing enterocolitis with area-under-the-curve values above 0.80. From an operational standpoint, automated ultrasound workflows deliver biometric measurements at about 14 milliseconds per frame, and dynamic scheduling in IVF laboratories lowers staff workload and per-cycle costs. Home-monitoring platforms for pregnant women are associated with 7–11 percent reductions in maternal mortality and preeclampsia incidence. Despite these advances, most evidence derives from retrospective, single-center studies with limited external validation. Low-resource settings, especially in Sub-Saharan Africa, remain under-represented, and few AI solutions are fully embedded in electronic health records. Conclusions: AI holds transformative promise for perinatal care but will require prospective multicenter validation, equity-centered design, robust governance, transparent fairness audits, and seamless electronic health record integration to translate these innovations into routine practice and improve maternal and neonatal outcomes. Full article

This paper shows that artificial intelligence is fundamentally transforming college curricula by enabling data-driven personalization, which enhances student outcomes and better aligns educational programs with evolving workforce demands. Specifically, predictive analytics, machine learning algorithms, and natural language processing were applied here, grounded in constructivist learning theory and Human–Computer Interaction principles, to evaluate student performance and identify at-risk students to propose personalized learning pathways. Results indicated that the AI-based curriculum achieved much higher course completion rates (89.72%) as well as retention (91.44%) and dropout rates (4.98%) compared to the traditional model. Sentiment analysis of learner feedback showed a more positive learning experience, while regression and ANOVA analyses proved the impact of AI on enhancing academic performance to be real. Therefore, the learning content delivery for each student was continuously improved based on individual learner characteristics and industry trends by AI-enabled recommender systems and adaptive learning models. Its advantages notwithstanding, the study emphasizes the need to address ethical concerns, ensure data privacy safeguards, and mitigate algorithmic bias before an equitable outcome can be claimed. These findings can inform institutions aspiring to adopt AI-driven models for curriculum innovation to build a more dynamic, responsive, and learner-centered educational ecosystem. Full article

Human-Robot Collaboration (HRC) is pivotal for flexible, worker-centric manufacturing in Industry 5.0, yet dynamic task allocation remains difficult because operator states—fatigue and skill—fluctuate abruptly. I address this gap with a hybrid framework that couples real-time perception and double-estimating reinforcement learning. A Convolutional Neural Network (CNN) classifies nine fatigue–skill combinations from synthetic physiological cues (heart-rate, blink rate, posture, wrist acceleration); its outputs feed a Double Deep Q-Network (DDQN) whose state vector also includes task-queue and robot-status features. The DDQN optimises a multi-objective reward balancing throughput, workload and safety and executes at 10 Hz within a closed-loop pipeline implemented in MATLAB R2025a and RoboDK v5.9. Benchmarking on a 1000-episode HRC dataset (2500 allocations·episode⁻¹) shows the hybrid CNN+DDQN controller raises throughput to 60.48 ± 0.08 tasks·min⁻¹ (+21% vs. rule-based, +12% vs. SARSA, +8% vs. Dueling DQN, +5% vs. PPO), trims operator fatigue by 7% and sustains 99.9% collision-free operation (one-way ANOVA, p < 0.05; post-hoc power 1 − β = 0.87). Visual analyses confirm responsive task reallocation as fatigue rises or skill varies. The approach outperforms strong baselines (PPO, A3C, Dueling DQN) by mitigating Q-value over-estimation through double learning, providing robust policies under stochastic human states and offering a reproducible blueprint for multi-robot, Industry 5.0 factories. Future work will validate the controller on a physical Doosan H2017 cell and incorporate fairness constraints to avoid workload bias across multiple operators. Full article

Federated learning (FL) struggles under non-IID client data when local models drift toward conflicting optima, impairing global convergence and performance. We introduce entropy-regularized federated optimization (ERFO), a lightweight client-side modification that augments each local objective with a Shannon entropy penalty on the per-parameter update distribution. ERFO requires no additional communication, adds a single-scalar hyperparameter $\lambda$, and integrates seamlessly into any FedAvg-style training loop. We derive a closed-form gradient for the entropy regularizer and provide convergence guarantees: under $\mu$-strong convexity and L-smoothness, ERFO achieves the same $O(1/T)$ (or linear) rates as FedAvg (with only $O\left(\lambda \right)$ bias for fixed $\lambda$ and exact convergence when ${\lambda }_t\to 0$); in the non-convex case, we prove stationary-point convergence at $O(1/T)$. Empirically, on five-client non-IID splits of the UNSW-NB15 intrusion-detection dataset, ERFO yields a +1.6 pp gain in accuracy and +0.008 in macro-F1 over FedAvg with markedly smoother dynamics. On a three-of-five split of PneumoniaMNIST, a fixed $\lambda$ matches or exceeds FedAvg, FedProx, and SCAFFOLD—achieving 90.3% accuracy and 0.878 macro-F1—while preserving rapid, stable learning. ERFO’s gradient-only design is model-agnostic, making it broadly applicable across tasks. Full article

Background/Objectives: Activities of Daily Living (ADLs) are crucial for assessing an individual’s autonomy, encompassing tasks such as eating, dressing, and moving around, among others. Predicting these activities is part of health monitoring, elderly care, and intelligent systems, improving quality of life, and facilitating early dependency detection, all of which are relevant components of personalized health and social care. However, the automatic classification of ADLs from sensor data remains challenging due to high variability in human behavior, sensor noise, and discrepancies in data acquisition protocols. These challenges limit the accuracy and applicability of existing solutions. This study details the modeling and evaluation of real-time ADL classification models based on batch learning (BL) and stream learning (SL) algorithms. Methods: The methodology followed is the Cross-Industry Standard Process for Data Mining (CRISP-DM). The models were trained with a comprehensive dataset integrating 23 ADL-centric datasets using accelerometers and gyroscopes data. The data were preprocessed by applying normalization and sampling rate unification techniques, and finally, relevant sensor locations on the body were selected. Results: After cleaning and debugging, a final dataset was generated, containing 238,990 samples, 56 activities, and 52 columns. The study compared models trained with BL and SL algorithms, evaluating their performance under various classification scenarios using accuracy, area under the curve (AUC), and F1-score metrics. Finally, a mobile application was developed to classify ADLs in real time (feeding data from a dataset). Conclusions: The outcome of this study can be used in various data science projects related to ADL and Human activity recognition (HAR), and due to the integration of diverse data sources, it is potentially useful to address bias and improve generalizability in Machine Learning models. The principal advantage of online learning algorithms is dynamically adapting to data changes, representing a significant advance in personal autonomy and health care monitoring. Full article

Nowadays, more enterprises are rapidly transitioning to cloud computing as it has become an ideal platform to perform the development and deployment of software systems. Because of its growing popularity, around ninety percent of enterprise applications rely on cloud computing solutions. The inherent dynamic and uncertain nature of cloud computing makes it difficult to accurately measure the exact state of a system at any given point in time. Potential challenges arise with respect to task scheduling, load balancing, resource allocation, governance, compliance, migration, data loss, and lack of resources. Among all challenges, task scheduling is one of the main problems as it reduces system performance due to improper utilization of resources. State Action Reward Action (SARSA) learning, a policy variant of Q learning, which learns the value function based on the current policy action, has been utilized in task scheduling. But it lacks the ability to provide better heuristics for state and action pairs, resulting in biased solutions in a highly dynamic and uncertain computing environment like cloud. In this paper, the SARSA learning ability is enriched by the guidance of the Large Language Model (LLM), which uses LLM heuristics to formulate the optimal Q function. This integration of the LLM and SARSA for task scheduling provides better sampling efficiency and also reduces the bias in task allocation. The heuristic value generated by the LLM is capable of mitigating the performance bias and also ensuring the model is not susceptible to hallucination. This paper provides the mathematical modeling of the proposed LLM_SARSA for performance in terms of the rate of convergence, reward shaping, heuristic values, under-/overestimation on non-optimal actions, sampling efficiency, and unbiased performance. The implementation of the LLM_SARSA is carried out using the CloudSim express open-source simulator by considering the Google cloud dataset composed of eight different types of clusters. The performance is compared with recent techniques like reinforcement learning, optimization strategy, and metaheuristic strategy. The LLM_SARSA outperforms the existing works with respect to the makespan time, degree of imbalance, cost, and resource utilization. The experimental results validate the inference of mathematical modeling in terms of the convergence rate and better estimation of the heuristic value to optimize the value function of the SARSA learning algorithm. Full article

The proliferation of 5G networks has revolutionized wireless communication by delivering enhanced speeds, ultra-low latency, and widespread connectivity. However, in heterogeneous cloud radio access networks (H-CRAN), efficiently managing inter-cell interference while ensuring energy conservation remains a critical challenge. This paper presents a novel energy-efficient, dynamic enhanced inter-cell interference coordination (eICIC) scheme based on deep reinforcement learning (DRL). Unlike conventional approaches that focus primarily on optimizing parameters such as almost blank subframe (ABS) ratios and bias offsets (BOs), our work introduces the transmission power during ABS subframes (TPA) and the channel quality indicator (CQI) threshold of victim user equipments (CTV) into the optimization process. Additionally, this approach uniquely integrates energy consumption into the scheme, addressing both performance and sustainability concerns. By modeling key factors such as signal-to-interference-plus-noise ratio (SINR) and service rates, we introduce the concept of energy-utility efficiency to balance energy savings with quality of service (QoS). Simulation results demonstrate that the proposed scheme achieves up to $70\%$ energy savings while enhancing QoS satisfaction, showcasing its potential to significantly improve the efficiency and sustainability of future 5G H-CRAN deployments. Full article

Sea surface salinity (SSS) observed by satellite has been widely used since the successful launch of the first salinity satellite in 2009. However, compared with other oceanographic satellite products (e.g., sea surface temperature, SST) that became operational in the 1980s, the SSS product is less mature and lacks effective validation from the user end. We employed an unsupervised machine learning approach to classify the Level 3 SSS bias from the Soil Moisture Active Passive (SMAP) satellite and its observing environment. The classification model divides the samples into fifteen classes based on four variables: satellite SSS bias, SST, rain rate, and wind speed. SST is one of the most significant factors influencing the classification. In regions with cold SST, satellite SSS has an accuracy of less than 0.2 PSU (Practical Salinity Unit), mainly due to the higher uncertainty in the cold environment. A small number of observations near the seawater freezing point show a significant fresh bias caused by sea ice. A systematic bias of the SMAP SSS product is found in the mid-latitudes: positive bias tends to occur north (south) of 45°N(S) and negative bias is more common in 25°N(S)–45°N(S) bands, likely associated with the SMAP calibration scheme. A significant bias also occurs in regions with strong ocean currents and eddy activities, likely due to spatial mismatch in the highly dynamic background. Notably, satellite SSS and in situ data correlations remain good in similar environments with weaker ocean dynamic activities, implying that satellite salinity data are reliable in dynamically active regions for capturing high-resolution details. The features of the SMAP SSS shown in this work call for careful consideration by the data user community when interpreting biased values. Full article

Research in reinforcement learning indicates that animals respond differently to positive and negative reward prediction errors, which can be calculated by assuming learning rate bias. Many studies have shown that humans and other animals have learning rate bias during learning, but it is unclear whether and how the bias changes throughout the entire learning process. Here, we recorded the behavior data and the local field potentials (LFPs) in the striatum of five pigeons performing a probabilistic learning task. Reinforcement learning models with and without learning rate biases were used to dynamically fit the pigeons’ choice behavior and estimate the option values. Furthemore, the correlation between the striatal LFPs power and the model-estimated option values was explored. We found that the pigeons’ learning rate bias shifted from negative to positive during the learning process, and the striatal Gamma (31 to 80 Hz) power correlated with the option values modulated by dynamic learning rate bias. In conclusion, our results support the hypothesis that pigeons employ a dynamic learning strategy in the learning process from both behavioral and neural aspects, providing valuable insights into reinforcement learning mechanisms of non-human animals. Full article

Previous literature on deep learning theory has focused on implicit bias with small learning rates. In this work, we explore the impact of data separability on the implicit bias of deep learning algorithms under the large learning rate. Using deep linear networks for binary classification with the logistic loss under the large learning rate regime, we characterize the implicit bias effect with data separability on training dynamics. From a data analytics perspective, we claim that depending on the separation conditions of data, the gradient descent iterates will converge to a flatter minimum in the large learning rate phase, which results in improved generalization. Our theory is rigorously proven under the assumption of degenerate data by overcoming the difficulty of the non-constant Hessian of logistic loss and confirmed by experiments on both experimental and non-degenerated datasets. Our results highlight the importance of data separability in training dynamics and the benefits of learning rate annealing schemes using an initial large learning rate. Full article

The key module for autonomous mobile robots is path planning and obstacle avoidance. Global path planning based on known maps has been effectively achieved. Local path planning in unknown dynamic environments is still very challenging due to the lack of detailed environmental information and unpredictability. This paper proposes an end-to-end local path planner n-step dueling double DQN with reward-based $ϵ$-greedy (RND3QN) based on a deep reinforcement learning framework, which acquires environmental data from LiDAR as input and uses a neural network to fit Q-values to output the corresponding discrete actions. The bias is reduced using n-step bootstrapping based on deep Q-network (DQN). The $ϵ$-greedy exploration-exploitation strategy is improved with the reward value as a measure of exploration, and an auxiliary reward function is introduced to increase the reward distribution of the sparse reward environment. Simulation experiments are conducted on the gazebo to test the algorithm’s effectiveness. The experimental data demonstrate that the average total reward value of RND3QN is higher than that of algorithms such as dueling double DQN (D3QN), and the success rates are increased by 174%, 65%, and 61% over D3QN on three stages, respectively. We experimented on the turtlebot3 waffle pi robot, and the strategies learned from the simulation can be effectively transferred to the real robot. Full article