<?xml version="1.0" encoding="utf-8"?>
    <rss version="2.0">
      <channel xmlns:content="http://purl.org/rss/1.0/modules/content/">
        <title>Frontiers in Artificial Intelligence | Machine Learning section | New and Recent Articles</title>
        <link>https://www.frontiersin.org/journals/artificial-intelligence/sections/machine-learning</link>
        <description>RSS Feed for Machine Learning section in the Frontiers in Artificial Intelligence journal | New and Recent Articles</description>
        <language>en-us</language>
        <generator>Frontiers Feed Generator,version:1</generator>
        <pubDate>2026-09-13T00:12:43.17+00:00</pubDate>
        <ttl>60</ttl>
        <item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1900160</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1900160</link>
        <title><![CDATA[NeuroTrustNet: a cost-effective multimodal ensemble framework for brain tumor classification under cross-dataset variability]]></title>
        <pubdate>2026-09-11T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Ferdaus Ibne Aziz</author><author>Insoo Koo</author><author>Tumennast Erdenebold</author><author>Jubayer A. Hossain</author><author>Asle Fagerstrøm</author><author>Debasish Ghose</author>
        <description><![CDATA[Accurate brain tumor classification from Magnetic Resonance Imaging (MRI) remains challenging due to dataset heterogeneity, class imbalance, and limited interpretability, while practical deployment further requires models that balance predictive performance with computational efficiency. In this work, we propose NeuroTrustNet, an integrated multimodal framework that combines complementary convolutional neural network (CNN), Vision Transformer (ViT), and handcrafted radiomic representations through the proposed Adaptive Attention Stacking (AAS) mechanism, enabling sample-specific feature fusion to improve robustness under cross-dataset variability while maintaining deployment-oriented computational efficiency. To evaluate performance under different computational constraints, we consider both high-capacity ensemble models (CNN and ViT ensembles) and lightweight architectures (RapidNet and AdaptoVision) as baseline systems. Experimental results on a large multi-dataset corpus show that while high-capacity ensembles achieve the highest accuracy (up to 96% on an external test set), the proposed NeuroTrustNet maintains competitive performance (94%) while reducing the computational cost of the fusion optimization stage by approximately 80% compared with full end-to-end ensemble training, highlighting its suitability for deployment-oriented medical AI systems operating under computational constraints while maintaining an effective balance between accuracy and efficiency. To further enhance interpretability, we incorporate a post-hoc interpretability component combining visual attribution maps with structured textual summaries derived from model outputs, enabling transparent and human-readable insights without influencing model predictions. The results demonstrate the potential of NeuroTrustNet as a deployment-oriented multimodal decision-support framework, providing a competitive balance between predictive performance, computational efficiency, and interpretability under cross-dataset variability.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1888120</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1888120</link>
        <title><![CDATA[MSAR: multi-stage adversarial representation learning for robust financial entity-level sentiment analysis]]></title>
        <pubdate>2026-09-11T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Xinwen Wan</author><author>Junjie Zhao</author><author>Zheng Liang</author><author>Aoxue Shen</author><author>Wenwen Chen</author><author>Andi Xiao</author>
        <description><![CDATA[Financial entity-level sentiment analysis aims to identify sentiment polarity toward specific financial entities in financial texts. This task is challenging because a single sentence may contain multiple entities with opposite sentiment orientations, while financial sentiment is often conveyed through domain-specific and implicit cues, increasing the risk of entity confusion. To address these challenges, we propose MSAR, a Multi-Stage Adversarial Representation Learning framework built on a FinBERT-CRF backbone. At the textual level, a Model-Sensitive Metric generates adversarial examples by jointly considering model sensitivity, semantic consistency, and linguistic fluency. At the representation level, Adaptive Adversarial Regularization dynamically adjusts the regularization strength according to gradient sensitivity and output-distribution variation. By integrating discrete textual perturbation with continuous representation regularization, MSAR enhances robust entity-level sentiment prediction. Experiments on FinEntity and SentFin show that MSAR outperforms representative baselines. Ablation, sensitivity, and low-resource experiments further confirm the effectiveness and robustness of the proposed components. These findings demonstrate that jointly optimizing textual and representation-level robustness improves financial entity-level sentiment analysis, particularly under low-resource conditions.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1924449</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1924449</link>
        <title><![CDATA[NBE-VLT-PFO: hybrid deep learning transformer architecture for joint estimation of lithium-ion batteries]]></title>
        <pubdate>2026-09-10T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>G. Tamizharasi</author><author>G. K. Rajini</author>
        <description><![CDATA[Lithium-ion batteries are utilized in electric vehicles (EVs), and accurate predictions of battery state of charge (SOC), state of health (SOH), and remaining useful life (RUL) are needed to ensure reliability, security, and durability of these batteries. This study provides a hybrid deep learning (DL) system, termed as Neural Basis Expansion Variational Ladder Transformer (NBE-VLT), integrating Polar Fox Optimization (PFO) for effective global hyperparameter tuning to produce accurate SOC, SOH, and RUL predictions from raw battery cycle data. The preprocessed battery cycle data were sent through a series of steps before the analysis to increase data features and reliability; these steps involved removing outliers, encoding features, normalizing the data, and aligning time. The Neural Basis Expansion (NBE) layer decomposes the battery's temporal signals using an adaptive basis function. The Variational Ladder Autoencoder (VLAE) generates hierarchical latent representations capable of learning and capturing degradation patterns and uncertainty information from battery data. In addition, a transformer encoder is used to learn long-range temporal dependencies, and an attention fusion mechanism combines features at different scales to produce accurate estimates of SOC, SOH, and RUL. The efficiency of the proposed model is shown through experimental evaluation conducted with the National Aeronautics and Space Administration (NASA) battery datasets, achieving SOC predictions with R2 of 0.9943, root mean squared error (RMSE) of 0.0225, mean absolute error (MAE) of 0.0164, and mean absolute percentage error (MAPE) of 1.64%. The SOH prediction achieved an R2 of 0.9606, RMSE of 0.0414, MAE of 0.0243, and MAPE of 2.43%, whereas RUL predictions achieved an R2 of 0.9936, RMSE of 0.0196, MAE of 0.0156, and MAPE of 1.56%. This indicates that the proposed framework, NBE-VLT-PFO, offers greater accuracy, robustness, and reliability for battery health forecasting.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1876996</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1876996</link>
        <title><![CDATA[Deep learning-based EV battery fault diagnosis using RBBMO with CAR-TATNET detection approach]]></title>
        <pubdate>2026-09-10T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>G. Tamizharasi</author><author>G. K. Rajini</author>
        <description><![CDATA[Ensuring the safety and dependability of power batteries has become a major concern due to the rapid expansion of electric vehicles (EVs), and fault detection has emerged as an essential approach for guaranteeing system stability. This study develops a Deep Learning (DL)-based defect prediction technique that combines an optimization algorithm and a Context-Aware Recurrent Neural Network with Adaptive Temporal Transformer (CAR-TATNet) to overcome these drawbacks. The aim is to reduce noise and improve the extraction of pertinent operating modes from EV battery signals. The dataset was first pre-processed by data cleaning and with Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN). Key time-frequency features linked to EV battery fault patterns were captured using Short-Time Fourier Transform (STFT). Inspired by the way magpies forage, the Red-Billed Blue Magpie Optimization (RBBMO) algorithm finds the most important traits for effective feature selection. Finally, the CAR-TATNet model was employed to model long-term dependencies in temporal sequences and adaptively respond to time-varying fault patterns. The developed methodology performs better than current methods, according to experimental validation in Python software, with a superior accuracy of 94.94%, providing a potential alternative for precise EV battery fault detection.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1942787</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1942787</link>
        <title><![CDATA[Unsupervised deep learning framework for early detection of wellbore trajectory deviation in drilling operations]]></title>
        <pubdate>2026-09-07T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Prabhat Singh</author><author>Bushitha Vickram</author><author>Annmaria Benny</author><author>Kadeeja Mariyam</author><author>Sudakshina Dan</author><author>R. Harishwaran</author><author>B. Naveen Kumar</author><author>M. Aslam Abdullah</author>
        <description><![CDATA[Wellbore trajectory deviation remains one of the major operational challenges encountered during directional and extended-reach drilling because even small departures from the planned well path can lead to poor reservoir placement, wellbore instability, increased non-productive time, and significant drilling costs. In most field operations, trajectory monitoring depends on periodic directional surveys together with threshold-based diagnostics. Although these methods are widely used, they often identify deviations only after they have become operationally noticeable, limiting the opportunity for timely corrective action. To overcome the limitations of conventional monitoring, the present study proposes an unsupervised deep learning framework capable of identifying the early onset of trajectory deviation by analyzing integrated well-log and geo-mechanical data without relying on labeled deviation events. The proposed framework combines Depth, Gamma Ray (GR), Shale Volume (Vsh), Resistivity, Sonic Transit Time (ΔT), P-wave Velocity (Vp), S-wave Velocity (Vs), Bulk Density, Calculated Density, Neutron Porosity (NPHI), Density Porosity (DPHI), and Poisson's Ratio to capture the lithological and mechanical characteristics that influence drilling behavior and trajectory stability. An LSTM Autoencoder (LSTM-AE) is employed to learn the normal temporal evolution of drilling parameters and identify anomalous behavior through reconstruction error. To complement the sequential learning capability of the autoencoder, a Graph Neural Network (GNN) is developed to represent the physical and geological relationships among the measured parameters, allowing complex multivariate interactions to be analyzed without requiring labeled datasets. The performance of both models is evaluated by comparing their ability to distinguish normal drilling behavior from progressively unstable operating conditions. The obtained results demonstrate that the LSTM-AE effectively learns the sequential characteristics of stable drilling and provides reliable early warning through changes in reconstruction error, whereas the GNN offers improved discrimination of trajectory-related anomalies by modeling the underlying relationships between geological and geo-mechanical variables. Collectively, these complementary approaches enableearlier recognition of developing trajectory deviations while reducing false alarms compared with conventional monitoring techniques. The findings demonstrate that integrating temporal sequence modeling with graph-based relational learning provides a practical and scalable solution for intelligent wellbore trajectory monitoring, supporting improved wellbore stability assessment, safer drilling operations, and more informed decision-making in complex subsurface environments.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1962295</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1962295</link>
        <title><![CDATA[Editorial: Advanced integration of large language models for autonomous systems and critical decision support]]></title>
        <pubdate>2026-09-07T00:00:00Z</pubdate>
        <category>Editorial</category>
        <author>I. de Zarzà</author><author>J. de Curtò</author><author>Carlos T. Calafate</author>
        <description></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1840721</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1840721</link>
        <title><![CDATA[A robust privacy-preserving federated framework for kidney CT image classification using transfer learning models]]></title>
        <pubdate>2026-09-04T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Sai Sri Hemantha Konala</author><author>Srinivas Koppu</author>
        <description><![CDATA[IntroductionKidney abnormalities, including cysts, tumors, and stones, are the most common renal disorders that can lead to severe complications such as chronic kidney disease or renal failure. Deep learning-based medical image analysis offers an effective approach for the accurate classification of kidney abnormalities, aiding the early diagnosis of renal disorders. However, its centralized training leads to inadequate privacy protection.MethodsConsidering the importance of ensuring individuals' data privacy, this study proposes a novel federated transfer learning framework for accurate classification of renal abnormalities using 12,446 kidney CT scan images and simultaneously preserves data privacy. CT scan images were preprocessed by resizing and normalization, followed by data augmentation techniques, including random rotations (±30°), horizontal flips, and color jitter, to address class imbalance and improve model generalization. Five pre-trained deep learning models such as MobileNetV2, EfficientNetV2-S, ResNet50, DenseNet121, and InceptionResNetV2 were trained across seven federated clients. Federated weighted averaging was employed for aggregation, and AES-256 encryption in CBC mode was applied to all model parameter transmissions between clients and the server.ResultsMobileNetV2 achieved the best performance, attaining 99.48% accuracy, 99.29% precision, 99.32% recall, 99.3% F1-score, 0.9999 AUC-ROC, and log loss of 0.0247. Cross-client validation produced an average accuracy of 98.85% with a generalization gap of only −0.0063, indicating strong generalization across client datasets.DiscussionThe proposed framework provides an effective balance between privacy preservation and communication efficiency, highlighting its potential for deployment in distributed clinical environments for kidney disease diagnosis.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1867313</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1867313</link>
        <title><![CDATA[Robust vision transformer adaptation for UAV imagery]]></title>
        <pubdate>2026-09-03T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Jagan Murugesan</author><author>P. Helen Vijitha</author><author>L. Jai Vinita</author><author>A. K. Parvathy</author><author>P. N. Jeipratha</author>
        <description><![CDATA[Vision Transformers (ViTs) perform well on clean aerial imagery but degrade sharply when deployed in post-earthquake UAV operations, where motion blur, dust haze, illumination variation, and sensor noise combine to produce what we term post-earthquake visual drift, a structured distributional shift that can render an otherwise capable model dangerously unreliable in the field. Retraining or conventional domain adaptation is not a realistic option under the latency, compute, and annotation constraints of active disaster response. This work presents a comprehensive empirical evaluation of lightweight label-free test-time adaptation (TTA) methods for improving the robustness of ViTs under such deployment conditions, systematically comparing consistency-based self-supervision (MEMO) and entropy minimization (TENT) across datasets with varying classification complexity. Consistent with existing lightweight TTA approaches, only the LayerNorm affine parameters are adapted during inference, while the backbone and attention weights remain frozen. On the UAV-TEBDE post-earthquake dataset, accuracy rises from 73.47% under severe drift (severity 0.3) to 89.18% using MEMO, corresponding to a recovery of 15.71 percentage points over the drifted baseline, without using any ground-truth labels or performing any retraining. On UAV-TEBDE, MEMO also achieved higher recovery than the SAR baseline evaluated in this study. Additional evaluation on AID (30-class aerial scene classification) and UC Merced (21-class land-use classification) shows consistent drift-induced degradation and adaptation-driven recovery across datasets of varying complexity. The cross-dataset evaluation further reveals that the relative effectiveness of lightweight TTA methods depends on the classification-space dimensionality. TENT-based entropy minimization outperforms consistency-based MEMO when the class count is large, while MEMO is the stronger choice in low-class settings. This interaction between method design and classification-space dimensionality has direct consequences for choosing TTA strategies in operational disaster assessment pipelines. Feature-space PCA analysis confirms that LayerNorm-only updates geometrically recalibrate internal representations, rather than simply correcting the model's output logits.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1904923</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1904923</link>
        <title><![CDATA[The impact of artificial intelligence with multi-head self-attention based deep learning model for students' academic performance monitoring]]></title>
        <pubdate>2026-09-03T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Mahmoud Ragab</author>
        <description><![CDATA[IntroductionEducational data mining has been applied to examine students' data collected from different educational organizations to forecast academic students' performance, which could assist them in accomplishing improved results in their upcoming courses.MethodsThis study presents the impact of artificial intelligence with multi-head self-attention on students' academic development using metaheuristic optimization algorithms (IAIMSA-SADMOA) model. The study aims to advance an effective method for students' academic performance using artificial intelligence tools to improve learning outcomes, engagement, and personalized education. The min-max normalization is employed in the data normalization phase for transforming input data into a beneficial format. The fruit fly optimization algorithm is deployed for the feature selection process to select the most related features from a dataset. Moreover, the proposed model designs bidirectional long short-term memory and multi-head self-attention mechanisms for the student's academic performance classification process. The parameter tuning process is performed through a sparrow search algorithm to enhance the classification performance.Results and DiscussionThe experimental evaluation of the IAIMSA-SADMOA technique occurs using a benchmark dataset. The empirical results indicated the enhanced performance of the proposed method in comparison with recent approaches.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1832131</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1832131</link>
        <title><![CDATA[Advanced security framework for 5G-SDN attack detection and mitigation using Affinity Cohesive Clustering and Improved Chaotic Walrus-Optimized Convolutional LSTM with Deep Dual Info-Varcoder]]></title>
        <pubdate>2026-09-03T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Shameli R.</author><author>Sujatha Rajkumar</author>
        <description><![CDATA[IntroductionThe integration of 5G and Software-Defined Networking (SDN) has introduced new security challenges due to limited processing capabilities, inefficient resource allocation, frequent user mobility, and the rapid growth of Internet of Things (IoT) devices. These dynamic network conditions increase the likelihood of malicious activities, highlighting the need for robust and intelligent intrusion detection techniques.MethodsThis research introduces a hybrid intrusion detection framework, termed APCHC-DDST-IV-CLSTM-ICWO, which integrates Affinity Propagated Cohesive Hierarchical Clustering, Deep Dual Stream SpaTemp Info-Varcoder, Dense Convolutional Long Short-Term Memory (CLSTM), and Improved Chaotic Walrus Optimization. The clustering module identifies attack patterns, while the dual-stream Info-Varcoder extracts representative spatial and temporal features. The Dense CLSTM captures sequential dependencies in network traffic to improve attack classification, and the Improved Chaotic Walrus Optimisation algorithm optimises the model parameters to enhance predictive performance. The proposed approach was validated using a 5G-SDN dataset generated with the MATLAB SDN Toolbox, with network traffic and attack scenarios simulated through the NS-3 network simulator.ResultsThe model showed a superior classification performance with a high average accuracy of 0.985±0.002, precision of 0.984±0.003, and F1 score of 0.985±0.003. The model’s low log-loss of 0.411 allows for the accurate separation of attack categories, including unknown attacks. This method improves the security and performance of 5G-SDN by increasing the throughput to 3218 samples/second and the detection time to 1.74 seconds.DiscussionThe results of the experiment demonstrate that the proposed APCHC-DDST-IV-CLSTM-ICWO framework offers a feasible method to secure 5G-SDN networks. The system provides high detection accuracy, effective processing, and reliable identification of emerging cyber threats by combining hierarchical clustering, deep feature learning, sequence modeling, and optimization algorithms. These findings indicate that it is appropriate for improving the operational effectiveness and security of next-generation 5G-SDN configurations.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1903342</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1903342</link>
        <title><![CDATA[Multi-resolution perceiver-based residual attention for explainable fault diagnosis in power transmission systems]]></title>
        <pubdate>2026-09-03T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Shantanu Dinesh Wani</author><author>S. Abinaya</author><author>S. Abirami</author><author>R. Priyadarshini</author>
        <description><![CDATA[Fault diagnosis in power transmission systems requires the simultaneous identification of fault type and location from high-dimensional, multichannel, and temporally evolving measurements. Conventional deep learning approaches may struggle to jointly capture localized transient patterns, global temporal dependencies, and spatially distributed fault signatures while providing interpretable diagnostic information. To address these challenges, a Multi-Resolution Perceiver-Based Residual Attention (MRP-RA) framework is proposed, that combines temporal convolutional feature extraction with a learnable frequency-aware projection branch, Perceiver-based latent modeling, residual spatio-temporal attention, and coupled multi-task learning for fault detection, classification, and localization. The framework processes 256-step windows of 42-channel measurements from a simulated IEEE 5-bus transmission system, where the temporal and frequency-aware representations are fused before being compressed into a latent representation using 32 learnable tokens. Residual temporal and channel attention mechanisms are employed to identify salient temporal regions and feature channels associated with fault signatures, while the multi-task objective jointly optimizes anomaly detection, fault-type classification, and fault localization. Experimental results show that MRP-RA achieves 96.47% fault-type classification accuracy and 97.64% fault-location accuracy across multiple random seeds, while achieving 3.5 × higher inference throughput than the Transformer baseline under the evaluated configuration. These results demonstrate that the proposed combination of multi-resolution feature extraction, latent global modeling, and residual attention can provide accurate joint fault diagnosis with reduced attention-related computational requirements and interpretable attention-based diagnostic insights. The findings support MRP-RA as a promising approach for explainable fault diagnosis in simulated power transmission systems, while further validation across different grid topologies, operating conditions, and real-world measurements is required to establish broader generalizability.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1871926</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1871926</link>
        <title><![CDATA[Accuracy–latency trade-offs and absent embedded validation in deep learning driver drowsiness detection: a PRISMA systematic review]]></title>
        <pubdate>2026-08-31T00:00:00Z</pubdate>
        <category>Systematic Review</category>
        <author>Christian Wilbert Salas Yupanqui</author><author>Cristhian Edy Llanque Tipo</author><author>Frank Diego Choquehuanca Huayhua</author><author>Angel Rosendo Condori-Coaquira</author><author>David Mamani-Pari</author><author>Milton Edward Humpiri-Flores</author><author>Esteban Tocto-Cano</author>
        <description><![CDATA[IntroductionDriver drowsiness is a leading cause of road traffic fatalities worldwide, and the convergence of computer vision and deep learning has transformed driver-state monitoring by enabling non-invasive detection within Advanced Driver Assistance Systems (ADAS).MethodsThis systematic literature review, conducted using the PRISMA 2020 methodology and the Kitchenham protocol, synthesizes findings from 33 peer-reviewed studies published between January 2021 and October 2025 to examine the architectures, biomarkers, performance benchmarks, and deployment barriers of deep-learning-based drowsiness detection.ResultsThe literature is dominated by convolutional single–pass designs; no reviewed model combines an accuracy above 99% with a latency below 100 ms, and none were evaluated on embedded or automotive–grade hardware.DiscussionThis distribution highlights a persistent accuracy-latency trade-off and emphasizes the need for embedded hardware and cross–dataset evaluation. By synthesizing the available evidence, this review characterizes the current state of the art and proposes a prioritized framework for selecting deep learning architectures for embedded ADAS applications.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1868885</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1868885</link>
        <title><![CDATA[Benchmark-grounded self-supervised contrastive learning for AMI-based distribution grid monitoring]]></title>
        <pubdate>2026-08-31T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Ling Zhang</author><author>Bowei Yang</author><author>Xingsi Ke</author><author>Hong Zhao</author><author>Wenhua Zhang</author><author>Yumin Yao</author>
        <description><![CDATA[IntroductionSelf-supervised contrastive learning is effective for time-series representation learning, but its deployment in industrial environments remains challenging because of spatiotemporal heterogeneity, domain-specific noise, and label scarcity. We study these challenges in the smart-grid domain, a representative cyber-physical system characterized by non-stationary events, missing data, and policy-driven distribution shifts.MethodsWe introduce the West China AMI Benchmark (WCA-Bench), a real-world dataset containing 713,700 user-day electricity profiles from 1,950 users and 506 expert-verified structural evolution events, including electric-vehicle access and photovoltaic installation. We further propose Domain-Aware Self-Supervised Contrastive Learning (DA-SSCL), which combines a lightweight PatchMLP encoder, physically grounded augmentations, and a hierarchical instance- and patch-level InfoNCE objective.ResultsDA-SSCL achieves an unsupervised clustering ARI of 0.699, reaches 88.4% classification accuracy with only 10% labeled data, and maintains transferability across the evaluated heterogeneous city districts.DiscussionThese results provide a practical methodology and a verified benchmark for robust, label-efficient, and transferable time-series representation learning in cyber-physical systems.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1882410</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1882410</link>
        <title><![CDATA[Deep generative modeling for AI-guided inverse design of perovskite photovoltaic devices]]></title>
        <pubdate>2026-08-28T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Parvez Amin Khan</author><author>Muhammad Tipu Sultan</author><author>Md Mahamudul Islam</author><author>Md. Emran Hossain</author><author>Samiur Rahman</author>
        <description><![CDATA[IntroductionPerovskite solar cells (PSCs) have rapidly approached the performance ceiling of mature single-junction photovoltaics, yet further improvement is constrained by the high-dimensional, non-linear coupling between device parameters and power-conversion efficiency (PCE). This work presents an end-to-end AI-guided inverse-design framework that learns the conditional distribution of device parameters given target photovoltaic figures of merit.MethodsThe framework is trained and validated on 49,998 drift-diffusion simulations of PSCs balanced across three classes of dominant recombination mechanism. A physics-informed feature-engineering pipeline feeds an ensemble of forward surrogate models under a strictly leakage-controlled five-fold cross-validation protocol. A conditional variational autoencoder with feature-wise linear modulation (FiLM) and classifier-free guidance (CFG) generates device candidates conditioned on target Voc, Jsc, FF and PCE.ResultsThe XGBoost surrogate achieves R2 = 0.8661 ± 0.0020 on the PCE proxy, statistically outperforming five competitors (Wilcoxon p < 10−190) while indistinguishable from LightGBM (p = 0.51). SHAP, permutation importance, and mutual-information converge on parasitic series resistance and grain-boundary defect density as dominant PCE-limiting parameters. At the optimal guidance scale (w = 1.5), cVAE+CFG achieves hit-rates of 73.7%, 12.6%, and 3.4% at the 90th-, 99th-percentile and “Ultra” targets—improvements of 8.9 ×, 25.2 ×, and ≥34 × over random sampling, with 100% valid/unique and ≥99.8% novel candidates.DiscussionKolmogorov-Smirnov tests confirm generated devices preserve energy-level marginals while concentrating mass in the high-performance sub-manifold. The framework offers a transferable, reproducible, and statistically rigorous methodology for accelerating design of next-generation perovskite PV devices.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1908848</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1908848</link>
        <title><![CDATA[Deep learning techniques for extreme rainfall prediction: a review of state-of-the-art]]></title>
        <pubdate>2026-08-27T00:00:00Z</pubdate>
        <category>Review</category>
        <author>Braiton U. Mukhalela</author><author>Serestina Viriri</author><author>David Ndzi</author><author>Michael Gebreslasie</author><author>Muhammad Zeeshan Shakir</author><author>Naeem Ramzan</author><author>Fahad Ayaz</author><author>Mary Lynch</author><author>Llinos Haf Spencer</author><author>Saloshni Naidoo</author><author>Nisha Nadesan-Reddy</author><author>Ozayr Mahomed</author><author>Natalie Dickinson</author><author>Fiona Henriquez</author>
        <description><![CDATA[Extreme rainfall events (ERES) are among the most challenging hydro-meteorological phenomena to forecast because the complex, nonlinear atmospheric processes involved span multiple spatial and temporal scales and are further intensified by rising climate variability. While Numerical Weather Prediction (NWP) models offer physically consistent representations of atmospheric dynamics, they still struggle to resolve localized convection, rapidly evolving storm systems, and rare, high-intensity precipitation events. This study systematically reviews recent advances in Artificial Intelligence (AI) for extreme rainfall prediction published between 2020 and 2026, using a transparent literature search strategy, predefined screening criteria, quality assessment, and a structured literature review matrix. The review synthesizes evidence from a final evidence base of 139 studies across several dimensions, including deep learning (DL) methodologies, multimodal data fusion, forecasting horizons, operational deployment, uncertainty quantification, and emerging intelligent forecasting paradigms. The analysis reveals a transition from deterministic DI models to hybrid, physics-informed, uncertainty-aware, and operational forecasting systems that increasingly integrate AI with physical knowledge and heterogeneous environmental observations. Despite notable progress, persistent challenges remain, including data scarcity, class imbalance, model generalization, physical consistency, uncertainty estimation, transferability, and the lack of standardized evaluation frameworks. This review provides a unified synthesis of current methodologies. It identifies key research gaps, highlighting the need for evaluation frameworks that address physical plausibility, uncertainty quantification, reliability, transferability, and operational relevance. The findings indicate that future extreme rainfall prediction systems should advance beyond accuracy-focused evaluation toward integrated, trustworthy, uncertainty-aware, and physically consistent forecasting approaches.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1819095</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1819095</link>
        <title><![CDATA[Anomaly detection method for business processes based on enhanced complex scenario logs]]></title>
        <pubdate>2026-08-27T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Yang Song</author><author>Ruoyuan Zhang</author>
        <description><![CDATA[The quality of logs is crucial for the reliability of business process anomaly detection. Deployed information systems are limited by hardware and cannot generate high-quality event logs containing complex scenario information, thus requiring post-enhancement of log quality. During this process, matching the same entities across different modalities is vital. To maximize the matching degree between complex scenarios and text, we propose a two-stage method for enhancing complex scenario logs. In the first stage, cross-modal entity alignment based on image multi-feature mapping using a Vision-Language Pretraining (VLP) model is employed. Local features are extracted from images, and a gated multi-fusion module is used to dynamically fuse the global features and local multi-features of the image. The Information Noise Contrastive Estimation (InfoNCE) loss function is utilized to train and optimize the model, ensuring tight alignment between images and text. Additionally, a multi-feature enhancer is designed to match the same entities between images and text, integrating image feature information into the original logs. In the second stage, a scenario simulation graph model is constructed to effectively fuse discrete sensor data in complex scenarios with logs, supplementing the discrete data that is often overlooked in the scenarios. Finally, the performance of anomaly detection on enhanced event logs is evaluated using eight anomaly detection models on real scenarios and six commonly used event log datasets. The results show that, compared with other event log datasets, enhanced event logs have higher accuracy in business process anomaly detection. Moreover, ablation experiments are conducted on three different levels of enhanced scenario event logs to verify the effectiveness of the method.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1857934</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1857934</link>
        <title><![CDATA[Confidence-aware pseudo-label selection and verifier training for semi-supervised LLM reasoning with minimal labels]]></title>
        <pubdate>2026-08-26T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Keizo Kato</author><author>Chenhui Chu</author><author>Yugo Murawaki</author><author>Sadao Kurohashi</author>
        <description><![CDATA[This paper studies how to improve the reasoning ability of large language models (LLMs) with minimal supervision. Recent gains in LLM reasoning largely come from learning intermediate reasoning traces, and many methods reduce supervision cost by using traces whose final answers are correct. In realistic settings, however, obtaining even those answer labels can be costly, motivating methods that extend reasoning from a very small labeled set with abundant unlabeled questions. Verifier-based semi-supervised learning is a promising approach: a verifier trained on a small labeled set can score reasoning traces on unlabeled questions and identify candidates for pseudo-labeling. However, even with a verifier, it remains unclear how pseudo-labeled samples should be selected to support downstream training. In particular, pseudo-label selection must balance quality and quantity. To address this, we introduce an adaptive threshold selection policy that chooses thresholds on validation data using pseudo-label precision and sample count. We further combine this policy with confidence-aware verifier training to support confidence-based selection. Experiments on verifiable math reasoning benchmarks show that, under our training setup, this combination improves downstream reasoning accuracy over the tested baselines and selects pseudo-labeled subsets with a more favorable reliability–coverage trade-off. These results suggest a practical design direction for verifier-guided pseudo-label selection in answer-verifiable, minimal-label reasoning settings.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1903125</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1903125</link>
        <title><![CDATA[CDMSP: a convolutional dense multi-scale pooling framework for low-light colonoscopy image enhancement]]></title>
        <pubdate>2026-08-26T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Mohan K.</author><author>Gopinath Palanisamy</author><author>Nisha J S.</author>
        <description><![CDATA[Colorectal cancer (CRC) remains one of the main cancer-related causes of death worldwide, and colonoscopy is an effective method for early identification of CRC and substantially reducing the risk. Colonoscopy images are often affected by non-uniform illumination and low-light conditions, which make it difficult to visualize mucosal textures and lesion margins. In this study, we introduce a low-light colonoscopy image enhancement framework, Convolutional Dense Attention Network (CDAN)-Dense- Multi-Scale Pooling (MSP) (CDMSP), to address this problem. The framework centers around a CDAN, which is constructed through dense connections and attention mechanisms to increase feature propagation and emphasize diagnostically important areas while reducing noise and lighting artifacts. In addition, a small MSP module is used to capture contextual information at different spatial scales, thereby helping preserve the details of the structures and edges. The overall loss function, which is a combination of structural similarity, perceptual similarity, edge-based contrast, and color fidelity, is used to maintain a balance between contrast enhancement and the preservation of anatomy. According to the experimental results, the proposed method achieved a Structural Similarity Index Measure (SSIM) of 0.8757, a Learned Perceptual Image Patch Similarity (LPIPS) of 0.0742, an edge-based contrast measure (EBCM) of 0.5601, and a Color Fidelity Index (CFI) of 95.1760. It is extremely efficient with an inference time of [23.3 milliseconds (ms), 42.99 frames per second (FPS)], making it suitable for real-time clinical use.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1846084</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1846084</link>
        <title><![CDATA[Machine learning-based solar radiation prediction across heterogeneous climatic zones using NASA POWER data: a case study in Nigeria]]></title>
        <pubdate>2026-08-26T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Oluwaseun Temitope Faloye</author><author>Grace Tina Awotoye</author><author>John Akinremi</author><author>Oluwatobi Solomon Olaleye</author><author>Ayoola Olamitomi Oluwadare</author><author>Oluwafemi E. Adeyeri</author><author>Laemthong Laokhongthavorn</author><author>Viroon Kamchoom</author>
        <description><![CDATA[Accurate prediction of solar radiation is essential for renewable energy planning, climate analysis, and agricultural productivity. Solar radiation data are required for the evaluation of solar power generation potential and crop water requirement computation, which is almost unavailable due to insufficient instrument for its measurement in many locations of the developing nations. Studies that applied several kernels of the Gaussian Process Regression (GPR) for the prediction of solar radiation are scarce. This study evaluates the performance of several machine learning models for solar radiation prediction across three climatic regions in Nigeria: Kano, Ibadan, and Onne. These locations represent distinct ecological zones ranging from the semi-arid savanna in northern Nigeria to the humid coastal rainforest in the southern region. Meteorological data used in this study were obtained from the NASA (National Aeronautics and Space Administration) POWER. Two machine learning approaches—Gaussian Process Regression (GPR) and Support Vector Machine (SVM)—were applied using different kernel configurations to predict solar radiation. Model performance was evaluated using standard statistical error metrics including Root Mean Square Error (RMSE), Mean Squared Error (MSE), Mean Absolute Error (MAE), Normalized Root Mean Square Error (NRMSE), and the coefficient of determination (R2). The results show that Gaussian Process Regression models generally produced the most accurate predictions across the study locations. In particular, the Exponential GPR model demonstrated the best performance in the humid regions of Ibadan and Onne, with RMSE values of 1.46 and 1.36 MJ.m2.d−1, respectively. Conversely, the Fine SVM model achieved the best prediction accuracy in Kano, recording the lowest RMSE value of 2.16 MJ.m2. d−1. These differences are attributed to variations in regional climatic conditions, where the relatively stable atmospheric conditions in the dry northern region favor SVM models, while the complex and highly variable atmospheric processes in humid regions are better captured by probabilistic models such as GPR. Overall, the findings highlight the importance of selecting appropriate machine learning models based on climatic characteristics when predicting solar radiation. The results provide useful insights for improving solar energy resource assessment and environmental modeling in Nigeria and similar tropical regions.]]></description>
      </item><item>
        <guid isPermaLink="true">https://www.frontiersin.org/articles/10.3389/frai.2026.1809361</guid>
        <link>https://www.frontiersin.org/articles/10.3389/frai.2026.1809361</link>
        <title><![CDATA[Uncertainty-aware federated temporal learning with explainable LLM-based coaching for privacy-preserving wearable health systems]]></title>
        <pubdate>2026-08-26T00:00:00Z</pubdate>
        <category>Original Research</category>
        <author>Duane Chembakassery</author><author>Harisankar R. Nair</author><author>J. Prassanna</author>
        <description><![CDATA[IntroductionThe growing use of wearable sensors enables continuous health and activity monitoring; however, challenges such as noisy data, privacy concerns, and device heterogeneity limit the effectiveness of centralized learning systems. This paper presents a privacy-first, AI-enhanced lifestyle coaching framework that integrates multi-stage signal cleaning, federated learning (FL), and explainable large language models (LLMs) with structured prompting for intelligent, decentralized health guidance. Unlike prior FL-based human activity recognition (HAR) systems, this work introduces (i) a quantitatively validated multi-stage denoising pipeline, (ii) local latent-space semantic imputation under federated constraints, and (iii) an explainable LLM-based coaching layer whose reasoning is explicitly grounded in aggregated federated activity representations rather than raw sensor data.MethodsThe proposed system employs a robust preprocessing pipeline comprising Hampel filtering, adaptive wavelet–Butterworth denoising, Variational Autoencoder (VAE)-based semantic imputation, and Kalman smoothing. The cleaned signals are used to train highly efficient Deep LSTM models on-device, and local updates are aggregated using a FedProx-based framework to mitigate client drift and preserve data privacy. To ensure rigorous evaluation and eliminate temporal data leakage, strict chronological data splitting is enforced alongside a 50% overlap stride constraint.ResultsThe multi-stage signal cleaning pipeline successfully achieved a 7–10 dB improvement in signal-to-noise ratio (SNR). Under the strict evaluation conditions, the global federated model established a highly realistic, leak-free baseline, achieving an average global accuracy of 68.08% and a Macro-F1 score of 0.60 on highly imbalanced, strictly unseen future time-series data. Furthermore, the LLM-based coaching module achieved a faithfulness score of 0.87.DiscussionThe high faithfulness of the coaching module proves that high-level, uncertainty-aware summaries are sufficient to generate personalized, transparent, and context-aware lifestyle recommendations without ever exposing raw sensor streams. Overall, the integration of uncertainty-aware signal preprocessing, federated temporal modeling, and explainable LLM-driven coaching establishes a mathematically sound, scalable, and secure architecture for next-generation AI-driven health systems.]]></description>
      </item>
      </channel>
    </rss>