Search Results (8,374)

A key requirement of associative learning studies is the ability to motivate the subject to acquire memory of the conditioned stimulus–unconditioned stimulus (CS–US) association. Although zebrafish have been found capable of acquiring CS–US associative memory, in many studies, the fish failed to learn. One reason for the failure, I argue in this perspective article, is that we do not yet know how to motivate zebrafish. I illustrate this problem using examples, and offer some solutions, based upon results obtained in my own laboratory for appetitive associative learning tasks for zebrafish. I highlight the value of considering the ethology and ecology of the zebrafish. I discuss why food may have been an ineffective US for zebrafish. I provide examples for how to improve the rewarding properties of food based upon the foraging behaviour of zebrafish in nature. I discuss the efforts to identify alternative USs, including the sight of conspecifics or the presence of other ecologically relevant stimuli. I theorize about conflicting motivators in zebrafish learning studies, including the effect of human handling versus that of experimenter-controlled USs. I conclude that systematic analyses of different USs are needed, along with detailed studies on how they may be optimized for the analysis of learning and memory in zebrafish. Full article

Background/Objectives: This study proposes a comprehensive classification framework for the automatic detection of Alzheimer’s disease using handwriting data. An enriched feature space is constructed by combining 18 baseline features extracted from raw handwriting signals with 30 additional features derived from established handwriting analysis studies, resulting in a total of 48 features. To enhance clinical practicality, a task reduction analysis is conducted by comparing the full dataset containing 25 handwriting tasks with a reduced dataset comprising 14 selected tasks. Methods: The proposed framework employs a two-stage evaluation strategy involving four feature selection methods (Random Forest Feature Importance, Extreme Gradient Boosting Feature Importance, L1 Regularization and Recursive Feature Elimination), three normalization techniques (Unnormalized, Min–Max and Z-Score), and five baseline machine learning classifiers (Random Forest, Logistic Regression, Multilayer Perceptron, XGBoost and Support Vector Machines). In the second stage, a dynamic ensemble learning strategy is introduced, where the most effective classifiers are adaptively selected for each cross-validation fold and integrated using soft and hard voting schemes. Results: The experimental results demonstrate that reducing the number of tasks leads to an improvement in average classification accuracy from 79.47% to 81.03%, while simultaneously decreasing training time and memory consumption by approximately 40% and 35%, respectively. The highest classification performance, achieving an accuracy of 94.20%, is obtained using the Hard Ensemble combined with L1-based feature selection. Conclusions: These findings highlight that the joint use of enriched feature representations, task reduction, and dynamic ensemble learning provides an effective and computationally efficient solution for handwriting-based Alzheimer’s disease detection. Full article

Understanding the coupling between hydrological dynamics and carbon sequestration is critical for predicting hemiboreal forest resilience to climate extremes. This study investigates water–carbon interactions in the Järvselja forest (Estonia) through a multi-objective hybrid modeling framework. We integrated long-term (2014–2025) eddy covariance flux measurements and daily meteorological data with a coupled architecture combining the process-based GR4J-Cemaneige model and a Long Short-Term Memory (LSTM) network. To validate the physical consistency of the deep learning component, we employed Support Vector Regression (SVR) diagnostic probes to map LSTM internal cell states against ERA5 soil moisture reanalysis data and in situ water table measurements. The combined LSTM + GR4J-Cemaneige model outperformed standalone approaches in the calibrated Reola catchment (NSE = 0.887), so by assuming hydrological similarity the hybrid model was regionalized to the streamflow ungauged Kalli basin. An in silico interpretability probe validated that the LSTM implicitly encoded physically meaningful soil moisture dynamics ($r>0.9$) without explicit training data. The analysis revealed that the 2018 heatwave triggered a synchronous collapse in water availability and carbon uptake, shifting the ecosystem from a robust sink to a net source. A significant legacy effect was observed, with carbon sequestration capacity lagging behind hydrological recovery for two years. The results of this paper substantiate the influence of climate warming on hemiboreal forests, demonstrating its implications for soil hydrology and the availability of water to sustain photosynthesis. Full article

The optimizer is a critical element of neural networks because it computes their optimal parameters through a training process. The Adam optimizer is considered the state of the art in deep learning. However, a drawback is the cost of storing and computing their gradients. A useful tool for addressing this issue is the application of the wavelet transform, and other relevant tool is the fractional derivative, which can be used to create fractional gradient optimizers. This research combines the wavelet transform and fractional optimizers to propose FAdamWav, a fractional version of Adam that uses (i) a parametric discrete wavelet transform to theoretically save $50\%$, $75\%$ or $87.5\%$ of gradient’s memory with one, two or three transformation levels, and (ii) a fractional gradient to optimize the neural network parameters. Experiments indicate that the saved memory is lower than the theoretical bounds, but memory is saved and fractional wavelet-based optimizers have competitive performance compared to their non-fractional and non-wavelet counterparts. Full article

In recent years, autonomous underwater vehicles (AUVs) have been increasingly employed for target surveillance and tracking. However, the limited performance and information-processing capability of a single AUV make it difficult to achieve high-precision tracking in practice. To address these challenges, this paper proposes an online-to-offline multi-agent reinforcement learning (MARL) framework that employs offline training on historical data to obtain the expert policy. Then, the optimal policy is generated by online fine-tuning technology, which enhances the training efficiency of reinforcement learning in new scenarios. To expand the surveillance range of AUV swarms, a distributed cooperative strategy based on area information entropy (AIE) is introduced. To reduce energy consumption in complex marine environments containing obstacles and vortices, ocean current and energy consumption models are introduced, together with an energy-efficiency optimization strategy. Furthermore, a long short-term memory (LSTM) network is integrated into the offline-to-online MARL framework to predict time-varying environmental states, thereby improving tracking accuracy and energy efficiency. Experimental results show that the proposed scheme is superior to the baseline schemes in terms of energy consumption, task success rate, and distance between AUVs. In addition, various performance indicators of the extended AUV swarm are also superior to the baseline schemes, demonstrating that the proposed scheme has excellent performance and scalability. Full article

This systematic review summarizes the application of artificial intelligence (AI) in classifying upper-limb motion using Electroencephalogram (EEG) and Electromyogram (EMG) signals, focusing on the field’s progression from Traditional Machine Learning (TML) to Deep Learning (DL) architectures. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, a search of PubMed, IEEEXplore, and Web of Science yielded 301 eligible studies published up to June 2025. The results indicate a change from classical classifiers like Linear Discriminant Analysis (LDA) and Support Vector Machines (SVMs) toward DL approaches. While Convolutional Neural Networks (CNNs) remain the most frequently implemented, emerging architectures, including Long Short-Term Memory (LSTM) networks and Transformers, have demonstrated remarkable performance. Despite the rise of DL, classical models remain highly relevant due to their robustness and efficiency. This review also identifies a heavy reliance on EEG-only modalities (60%), with only 7% of studies utilizing hybrid EEG-EMG systems, representing a potential missed opportunity for signal fusion. Full article

Human Activity Recognition (HAR) on wearable and IoT devices must jointly satisfy four requirements: high accuracy, the ability to recognize previously unseen activities, strict memory and latency constraints, and interpretable decisions. In this work, we address all four by introducing an explainable Kolmogorov–Arnold Network for Human Activity Recognition (TinyKAN-HAR) with a zero-shot learning (ZSL) module, designed specifically for TinyML edge devices. The proposed KAN replaces fixed activation functions by learnable one-dimensional spline operators applied after linear mixing, yielding compact yet expressive feature extractors whose internal nonlinearities can be directly visualized. On top of the KAN latent space, we learn a semantic projection and cosine-based compatibility function that align sensor features with class-level semantic embeddings, enabling both pure and generalized zero-shot recognition of unseen activities. We evaluate our method on three benchmark datasets (UCI HAR, WISDM, PAMAP2) under subject-disjoint and zero-shot splits. TinyKAN-HAR consistently achieves over 97% macro-F1 on seen classes and over 96% accuracy on unseen activities, with harmonic mean above 96% in the generalized ZSL setting, outperforming CNN, LSTM and Transformer-based ZSL baselines. For explainability, we combine gradient-based attributions, SHAP-style global relevance scores and inspection of the learned spline functions to provide sensor-level, temporal and neuron-level insights into each prediction. After 8-bit quantization and TinyML-oriented optimizations, the deployed model occupies only 145 kB of flash and 26 kB of RAM, and achieves an average inference latency of 4.1 ms (about 0.32 mJ per window) on a Cortex-M4F-class microcontroller, while preserving accuracy within 0.2% of the full-precision model. These results demonstrate that explainable, zero-shot HAR with near state-of-the-art accuracy is feasible on severely resource-constrained TinyML edge devices. Full article

Virtual reality offers unprecedented experiences and serves as a new learning tool. However, limited research has examined how this novel visual experience influences cognitive processes. This study described a conceptual theoretical framework of visual cognition, organizing cognition into four links: memory, analogy, logical thinking, and creative thinking. A between-subjects experiment was conducted with 54 college students, who were randomly assigned to a VR (N = 27) or a PC group (N = 27) to learn unfamiliar Mars-related content. Participants completed two learning sessions followed by cognitive tests after each session. Single-channel EEG was recorded at Fp1 during learning, together with proprietary attention and meditation metrics. EEG data were filtered, denoised using wavelet thresholding, and analyzed via short-time Fourier transform to calculate power spectral density and examine the differences in relative power. Results showed that the VR group performed significantly worse on second-session memory tests (p = 0.013). Besides, the VR group exhibited significantly lower meditation levels (p = 0.049) and reduced attention. EEG analysis revealed a significant decrease in the theta band (p = 0.010) and a significant increase in the beta band (p = 0.013). These findings suggest that, within the scope of this study, VR environments without prior knowledge may negatively affect attention, meditation, and memory performance, and provide preliminary evidence that VR is not automatically beneficial for novice learners. Full article

Background: The Perri Auditory Verbal Learning Test (Perri-AVLT) is a cognitive tool designed to assess verbal learning and memory. Currently, demographically adjusted norms for the Perri-AVLT are not available for elderly Mexican adults. Objective: This study aimed to develop regression-based norms from elderly Mexican adults to enable demographic adjustments for clinical interpretation. Methods: The sample included 294 elderly Mexican adults aged 60–89 (224 cognitively normal individuals, and 70 clinical cases) from Mexico (Jalisco, Guanajuato, and Mexico City). Participants were administered the Perri-AVLT. A multivariate regression-based norming approach was used to evaluate the effects of age, sex, and years of education on test performance. Results: The multivariate regression model showed that years of education were a significant predictor of cognitive performance across all Perri-AVLT trials. The Pearson correlation for all Perri-AVLT trials was high. Conclusion: This study provides regression-based normative data for the Perri-AVLT adjusted for sociodemographic factors. These norms can be used to evaluate verbal learning and memory in elderly Mexican adults. This information can support a neuropsychologist in cognitive assessment, rehabilitation, and research. Full article

It is proposed that the process of learning may be represented as a trajectory within the phase space of long-term memory. The research uses an artificial neural network design to explore, in theory, if starting from different points within the phase space affects how quickly learning occurs. Using a Monte Carlo method, 1000 virtual agents were trained using the Levenberg–Marquardt algorithm to recognise a large set of Arabic digits at ten different skill levels. The simulations replicated the typical learning curves observed in human learning and were successful in distinguishing ten levels of skill. First, and in line with previous research, the results provide convincing evidence that learning consolidates a selected set of pathways within the network. Second, and critical to the hypothesis, the distance in the phase space, calculated as the difference in average connectivity between skill levels, is highly predictive of both learning time and performance. The findings strongly support the hypothesis that learning represents progression along a trajectory connecting two points within the phase state landscape. As these properties may be more pronounced in biological systems because of their greater complexity, these results shed new light on individual variance in learning. Full article

Background: Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by social deficits, repetitive behaviors, and heightened anxiety. Despite extensive research, effective interventions targeting core symptoms remain limited. Gami-Guibitang (GBT), a traditional herbal formula, has been clinically prescribed for anxiety-related symptoms and cognitive complaints, yet its effects on ASD-associated behavioral and molecular abnormalities have not been fully elucidated. Objective: This study aimed to evaluate the anxiolytic and neuroregulatory effects of GBT in a valproic acid (VPA)-induced ASD mouse model, focusing on behavioral outcomes and hippocampal synaptic protein expression. Methods: Pregnant C57BL/6N mice received a single intraperitoneal injection of VPA (500 mg/kg) at embryonic day 12.5. Male offspring were administered GBT (150 mg/kg, p.o.) twice daily for 4 weeks from postnatal day 21 (PND 21). These mice were behaviorally evaluated by the open-field test, elevated plus maze, marble-burying test, Y-maze, three-chamber social interaction test, and Morris water maze. Western blot analysis was conducted to examine hippocampal expression of phosphorylated and total CREB and GluR1, PI3K/Akt signaling components, as well as GABRA1 and GABRB1. Results: VPA-exposed offspring exhibited increased anxiety-like behaviors, altered repetitive behaviors, dysregulated exploratory activity, and impaired spatial learning, and reduced spontaneous alternation performance in the Y-maze. GBT reduced anxiety-like behaviors in the elevated plus maze and marble burying tests, partially improved spatial learning acquisition in the Morris water maze, and normalized excessive locomotor activity, without significantly affecting short-term working memory performance. At the molecular level, GBT significantly attenuated VPA-induced hyperphosphorylation of CREB, GluR1, PI3K, and Akt, indicating suppression of aberrant synaptic signaling rather than global enhancement. In addition, GBT increased GABRA1 expression toward control levels and enhanced GABRB1 expression beyond baseline, suggesting selective modulation of GABAergic receptor subunit composition rather than simple normalization. Conclusions: These findings provide preclinical evidence that GBT alleviates anxiety-like behavior and modulates hippocampal synaptic signaling disrupted by prenatal VPA exposure. By attenuating aberrant excitatory signaling and selectively regulating GABAergic receptor balance, GBT may represent a multi-target herbal candidate for modulating ASD-associated emotional dysregulation and domain-specific cognitive dysfunction, rather than acting as a broad cognitive enhancer. Full article

In the post-Moore’s Law era, conventional Von Neumann architectures face critical limitations, such as the “memory wall” and excessive power consumption, particularly when processing unstructured data. Neuromorphic computing, inspired by the human brain, offers a promising solution through parallel processing and adaptive learning. Among the candidates for artificial synapses, memristors based on two-dimensional MXenes (specifically Ti₃C₂T_x) have attracted significant attention due to their unique layered structure, high metallic conductivity, and tunable physicochemical properties. This review provides a comprehensive analysis of MXene-based memristors, from material synthesis to system-level applications. We examine how different synthesis strategies, including etching methods, directly influence device performance and elucidate the underlying resistive switching mechanisms driven by ion migration, valence change, and interfacial processes. Furthermore, the review demonstrates the efficacy of MXenes in emulating biological synaptic functions—such as spike-timing-dependent plasticity (STDP) and long-term potentiation/depression (LTP/LTD)—and their application in tasks like handwritten digit recognition. Finally, we highlight emerging frontiers in flexible electronics and in-sensor computing, offering insights into the future trajectory of integrated sensing, memory, and computation. Full article

Price volatility in steel reinforcement bars (rebar) plays a pivotal role in managing construction project costs, with precise forecasting being essential for maintaining corporate profitability and ensuring market stability. This research conducts a comprehensive evaluation of five prominent forecasting models—Autoregressive Integrated Moving Average (ARIMA), eXtreme Gradient Boosting (XGBoost), Prophet, Long Short-Term Memory (LSTM), and Transformer—specifically applied to steel rebar price prediction. The study emphasizes the influence of feature selection, defined as the number of historical price data points utilized for prediction, on the accuracy of these models. Furthermore, it develops a hybrid forecasting framework grounded in a residual complementarity mechanism aimed at improving long-term predictive performance. The results reveal that the ARIMA model delivers consistent and reliable short-term forecasts, particularly within a two-month horizon, whereas the Prophet model effectively captures long-term price trends but suffers from notable short-term bias. A two-stage hybrid model (referred to as Combination Model II), which integrates ARIMA and Prophet through residual inversion, demonstrates superior forecasting accuracy over a six-month period. This hybrid approach surpasses the standalone ARIMA model by more than 70% across key evaluation metrics—including Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Symmetric Mean Absolute Percentage Error (SMAPE), and Mean Absolute Scaled Error (MASE)—and exceeds the performance of the standalone Prophet model by over 90%. This integration effectively combines the high short-term precision of ARIMA with the long-term trend stability of Prophet. Within the domain of machine learning and deep learning models, XGBoost achieves optimal predictive accuracy when utilizing between one and four features. The predictive performance of LSTM does not exhibit a straightforward linear relationship with the number of features; however, certain feature combinations enable it to outperform other models. Transformer models maintain stable accuracy when employing feature sets ranging from one to five and twelve to seventeen, but display considerable variability in performance when the feature count lies between five and twelve. This investigation delineates the optimal parameter ranges and contextual applicability for each model. The proposed hybrid forecasting methodology, alongside a model transfer strategy encompassing data preprocessing adjustments, parameter optimization, and weight adaptation, offers practical applicability to other commodity markets such as cement and concrete. Consequently, this research provides a scientifically grounded framework to support procurement decision-making processes within construction enterprises. Full article

Resource-constrained intelligent systems increasingly require reliable long-form reasoning capabilities under limited computational and memory budgets, particularly in edge and embedded sensing environments. However, reinforcement learning for long-horizon decision generation remains highly unstable in such low-resource settings due to severe reward sparsity and imbalanced credit assignment, which often lead to non-convergent or excessively verbose generation behavior. In this work, we propose AF-CuRL (Answer-Focused Curriculum Reinforcement Learning), a lightweight reinforcement learning framework designed to stabilize long-form generation without increasing model size or computational cost. AF-CuRL improves optimization learnability through two complementary objective-level designs: (1) answer-focused token reweighting, which concentrates policy updates on reward-critical regions of generated sequences to alleviate credit assignment imbalance, and (2) a two-phase curriculum reward schedule that prioritizes stable termination and output regularity before shifting toward correctness-oriented optimization. We evaluate AF-CuRL on a 1.5B-parameter language model under strictly constrained training settings, using mathematical reasoning tasks as a controlled and reproducible proxy for long-horizon, rule-based decision-making commonly encountered in intelligent sensing and embedded systems. Experimental results demonstrate consistent improvements in both decision accuracy and generation regularity, including higher termination reliability and reduced generation length, compared with standard sequence-level reinforcement learning baselines. These results suggest that, for resource-limited and edge-intelligent systems, structured objective design can be more effective than model scaling for achieving stable and efficient long-form reasoning, providing a practical reinforcement learning solution for intelligent systems operating under real-world constraints. Full article

Background/Objectives: Stewartia pseudocamellia Maxim. (S. pseudocamellia) has been reported to possess antioxidant and anti-inflammatory properties and contains various bioactive flavonoids and phenolic compounds. These components may contribute to neuroprotective effects relevant to depression and cognitive dysfunction. This study was conducted to evaluate the effects of 20% ethanolic extract from S. pseudocamellia leaves (ESP) on chronic unpredictable mild stress (CUMS)-induced depressive-like behaviors and cognitive dysfunction in C57BL/6 mice. Methods: C57BL/6 mice were divided into six groups: normal control (NC), normal sample (NS; ESP 100 mg/kg), CUMS, L-theanine (Thea; 4 mg/kg), ESP 50 mg/kg, and ESP 100 mg/kg groups. Phytochemical profiling of ESP was performed using ultra-performance liquid chromatography–quadrupole time-of-flight tandem mass spectrometry (UPLC-Q-TOF-MS/MS). Depressive-like behaviors and cognitive function were assessed, along with stress-related hormonal regulation and associated cellular signaling pathways. Results: Phytochemical profiling of ESP identified procyanidin B2, epicatechin, rutin, catechin gallate, kaempferol 3-O-glucoside, and quercitrin as major constituents. ESP significantly alleviated CUMS-induced depressive-like behaviors and improved spatial learning and memory. These effects were associated with modulation of stress-related hormones in serum and hypothalamic–pituitary–adrenal (HPA) axis–related proteins in the brain. ESP also enhanced antioxidant defense by activating the Nrf2 signaling pathway and improving mitochondrial function. Furthermore, ESP attenuated neuroinflammation and apoptosis by regulating the TLR4/NF-κB and JNK pathways, and promoted neuroplasticity by modulating cholinergic activity, with enhanced BDNF/TrkB signaling in the cerebral cortex and hippocampus. Conclusions: Collectively, these findings suggest that ESP exerts protective effects against CUMS-induced depressive-like behaviors and cognitive deficits in a preclinical model. Full article