IEEE transactions on artificial intelligence最新文献_第2页

Fast Clustering via Anchor Label Alignment and Transition Probabilities 基于锚标对齐和转移概率的快速聚类

IEEE transactions on artificial intelligence Pub Date : 2026-10-01 Epub Date: 2026-06-02 DOI: 10.1109/TAI.2026.3699280

Jingwei Chen;Jingqing Cheng;Shiyu Xie;Jia Zhao;Feiping Nie

{"title":"Fast Clustering via Anchor Label Alignment and Transition Probabilities","authors":"Jingwei Chen;Jingqing Cheng;Shiyu Xie;Jia Zhao;Feiping Nie","doi":"10.1109/TAI.2026.3699280","DOIUrl":"https://doi.org/10.1109/TAI.2026.3699280","url":null,"abstract":"Anchor-based clustering techniques have gained considerable attention for their efficiency and ability to capture representative information. However, traditional methods often suffer from separate stages of spectral feature extraction and discretization, leading to critical information loss and reduced clustering performance. Additionally, most existing models require tuning of hyperparameters introduced by various regularization terms, limiting their practicality. To address these issues, we propose a novel method that aligns anchor-to-anchor transition probabilities via the correlation of anchor labels, and implicitly incorporates objectives of maximizing intra-cluster similarity and balanced clustering, called fast clustering via anchor label alignment and transition probabilities (FC-ALATP). The regularization parameter does not require manual tuning, which be updated in closed form. Moreover, the optimization stage scales linearly with the number of anchors; specifically, two solvers, projected gradient descent (PGD) and fast coordinate descent (fast CD), are developed. Extensive experiments demonstrate that fast CD outperforms PGD, and FC-ALATP achieves superior performance over other anchor-based models with less time cost.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"7 10","pages":"5736-5752"},"PeriodicalIF":0.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097783","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

FMamba: Bi-Mamba Based on Fast Gated-Attention for Multivariate Time-Series Forecasting FMamba：基于快速门控注意的Bi-Mamba多元时间序列预测

IEEE transactions on artificial intelligence Pub Date : 2026-10-01 Epub Date: 2026-06-04 DOI: 10.1109/TAI.2026.3699838

Shusen Ma;Yun-Bo Zhao;Yu Kang

{"title":"FMamba: Bi-Mamba Based on Fast Gated-Attention for Multivariate Time-Series Forecasting","authors":"Shusen Ma;Yun-Bo Zhao;Yu Kang","doi":"10.1109/TAI.2026.3699838","DOIUrl":"https://doi.org/10.1109/TAI.2026.3699838","url":null,"abstract":"In multivariate time-series forecasting (MTSF), extracting both temporal features and intervariable correlations is crucial. While Transformer-based models excel at capturing variable correlations, their quadratic computational complexity creates significant overhead, particularly for inputs with numerous variables. Recently, Mamba, a selective state space model, has shown promise due to its linear complexity and strong temporal extraction capabilities. However, Mamba lacks the explicit capability to model intervariable dependencies found in Transformers, and its unidirectional nature limits its receptive field. To address these limitations while maintaining efficiency, we propose FMamba, a novel framework with global selective capabilities. FMamba integrates a bidirectional Mamba (Bi-Mamba) structure with Gated-Performers—an improved linear Transformer with a content-aware gating mechanism. Technically, Gated-Performers explicitly compute dependencies among input variables, while Bi-Mamba selectively models temporal dependencies. Experimental results on seven public datasets demonstrate that FMamba outperforms Mamba-based and Transformer-based baselines, reducing the mean squared error (MSE) by an average of 15.5% and 15.2%, respectively.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"7 10","pages":"5776-5789"},"PeriodicalIF":0.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097050","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A CVaR-Constrained Safe Reinforcement Learning Framework With Action Repair for Practical Portfolio Optimization 一种带有动作修复的cvar约束安全强化学习框架用于实际投资组合优化

IEEE transactions on artificial intelligence Pub Date : 2026-10-01 Epub Date: 2026-04-24 DOI: 10.1109/TAI.2026.3686783

Himanshu Choudhary;Arishi Orra;Manoj Thakur;Xiao-Zhi Gao;Prabhat Kumar Sahu

{"title":"A CVaR-Constrained Safe Reinforcement Learning Framework With Action Repair for Practical Portfolio Optimization","authors":"Himanshu Choudhary;Arishi Orra;Manoj Thakur;Xiao-Zhi Gao;Prabhat Kumar Sahu","doi":"10.1109/TAI.2026.3686783","DOIUrl":"https://doi.org/10.1109/TAI.2026.3686783","url":null,"abstract":"Portfolio optimization refers to allocating capital across multiple assets to attain an optimal balance between potential returns and associated risks under dynamic market conditions. While deep reinforcement learning (DRL) has shown significant promise in learning adaptive allocation strategies through continuous interaction with the market, most existing approaches struggle to incorporate practical investment constraints and robust risk measures. To address these limitations, this article introduces a constraint-aware and risk-sensitive DRL framework for portfolio optimization that operates within a constrained Markov decision process (CMDP) setting. The proposed model embeds conditional value-at-risk (CVaR) as a hard safety constraint using an augmented Lagrangian approach on the agent to prevent high-loss outcomes. Additionally, a repair mechanism is introduced to adjust the agent’s actions, which ensures the resulting portfolio weights remain feasible and compliant with real-world investment constraints. Our approach is evaluated using real-world stock data from multiple global indices and consistently outperforms existing DRL-based and traditional portfolio optimization strategies, achieving superior returns while strictly adhering to defined risk constraints.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"7 10","pages":"5417-5428"},"PeriodicalIF":0.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097157","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Diffusion-Based Sequential Recommender via Partial Noise Injection and Cross-Expert Comparison 基于局部噪声注入和跨专家比较的扩散序列推荐算法

IEEE transactions on artificial intelligence Pub Date : 2026-10-01 Epub Date: 2026-06-05 DOI: 10.1109/TAI.2026.3700980

Chunyuan Peng;Lin Zheng;Sentao Chen

{"title":"A Diffusion-Based Sequential Recommender via Partial Noise Injection and Cross-Expert Comparison","authors":"Chunyuan Peng;Lin Zheng;Sentao Chen","doi":"10.1109/TAI.2026.3700980","DOIUrl":"https://doi.org/10.1109/TAI.2026.3700980","url":null,"abstract":"In recent years, diffusion models have demonstrated remarkable performance across various domains and have shown their potential in recommender systems. However, in sequential recommendation, diffusion models face inherent limitations: the mismatch between the diffusion mechanism and sequence modeling exacerbates the negative impact of random noise on user preferences. In addition, user sequence information is not fully leveraged, leading to insufficient guidance in the forward and reverse processes and hindering the generation of high-quality preference representations. In this article, we propose a diffusion-based sequential recommender incorporating expert comparison and constraints to address these challenges. Specifically, we introduce a partial noise injection method to mitigate the impact of noise and enable conditional sequence guidance. Furthermore, by designing a cross-expert comparison and constraint mechanism, our approach effectively captures diverse sequential preferences, ultimately generating high-quality item representations. Experimental results demonstrate that our model produces representations that accurately reflect users’ multi-faceted preferences and achieves superior recommendation performance compared to state-of-the-art methods.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"7 10","pages":"5803-5812"},"PeriodicalIF":0.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097253","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

AffNet: Designing Multi-Headed Affinity and Adaptive Thresholding for Unsupervised Homophily Estimation 非监督同态估计的多头亲和和自适应阈值设计

IEEE transactions on artificial intelligence Pub Date : 2026-10-01 Epub Date: 2026-04-27 DOI: 10.1109/TAI.2026.3688210

Indranil Ojha;Kushal Bose;Swagatam Das

{"title":"AffNet: Designing Multi-Headed Affinity and Adaptive Thresholding for Unsupervised Homophily Estimation","authors":"Indranil Ojha;Kushal Bose;Swagatam Das","doi":"10.1109/TAI.2026.3688210","DOIUrl":"https://doi.org/10.1109/TAI.2026.3688210","url":null,"abstract":"The message-passing (MP) framework contributes to the profound success of the graph neural networks (GNNs). MP predominantly operates on homophilic graphs where the adjacent nodes are highly likely to have identical class labels. However, for heterophilic graphs containing adjacent nodes frequently with different class labels, numerous sophisticated algorithms have been proposed. A prior idea about the homophily of a graph helps choose the algorithm to be used. The estimation of homophily relies on the node labels which, practically, are often unavailable. Hence, we remain clueless about which GNNs should be applied to the input dataset. To mitigate the issue, we propose a novel and efficient method to measure the homophily without having prior information on the node labels. Our method employs trainable affinity vectors equipped with nodes apart from the pre-existing node features. The affinity vector of a node represents the feature vector that the node expects in a suitable neighbor. We measure affinity scores between nodes as the similarity between the affinity of one node to the features of another node. In this connection, we define a unique homophily measure based on the affinity scores of nodes with themselves. Extending the concept, the multi-headed affinity (MHA) vectors are designed to represent a diverse set of choices the nodes have to connect to other neighbors. Furthermore, we design an adaptive strategy to apply learned thresholds by going beyond the prevailing ones that use fixed thresholds. The estimated affinity scores are subsequently considered as the probability scores determining the existence of edges. In this context, we propose a robust architecture AffNet that can be used to simultaneously estimate homophily in an unsupervised setting, solve the link prediction task and build recommendation systems. Applying our framework to several standard and large-scale datasets, we obtain impressive performances compared with the various state-of-the-art methods.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"7 10","pages":"5512-5523"},"PeriodicalIF":0.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097333","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

GaitSS: Gait Slices as a Set for Gait Recognition in Interfered Sequences GaitSS：作为干扰序列步态识别的步态切片集

IEEE transactions on artificial intelligence Pub Date : 2026-10-01 Epub Date: 2026-04-27 DOI: 10.1109/TAI.2026.3687846

Huakang Li;Wing W. Y. Ng;Yidan Qiu;Ying Gao;Jianjun Zhang;Sam Kwong

{"title":"GaitSS: Gait Slices as a Set for Gait Recognition in Interfered Sequences","authors":"Huakang Li;Wing W. Y. Ng;Yidan Qiu;Ying Gao;Jianjun Zhang;Sam Kwong","doi":"10.1109/TAI.2026.3687846","DOIUrl":"https://doi.org/10.1109/TAI.2026.3687846","url":null,"abstract":"Video-based gait recognition suffers from environmental variations and deliberate disguises, which introduce information loss, redundancy, and complex temporal dependencies, ultimately degrading recognition performance. We propose GaitSS, a novel method that addresses these challenges by treating gait slices as a set rather than by maintaining a fixed spatial order within each frame. This treatment effectively mitigates the effects of misalignment, information loss, and redundancy. In addition, we introduce a temporal frame exchange augmentation (TFEA) strategy to enhance the model’s ability to capture robust temporal gait features. We introduce the GOSI dataset, specifically collected to include diverse interference conditions, such as varying lighting, camera heights, and intentional disguises. We also evaluate our method on three widely used public gait datasets: CASIA-B, OU-MVLP, and GREW, providing a comprehensive assessment. Our method outperforms existing methods, yielding rank-1 accuracies (averaged across conditions) of 94.3%, 92.4%, 80.3%, and 42.3% on the CASIA-B, OU-MVLP, GREW, and GOSI gait datasets, respectively. (42.3% is the average of NM, WH, and OC). Class activation maps show that GaitSS focuses more on discriminative features and less on interference regions, validating its superior handling of environmental variations. Ablation and validation experiments show that performance improvement is primarily due to treating gait slices as a set, rather than from image augmentation. GaitSS can be seamlessly integrated into existing gait recognition frameworks without introducing extra parameters or computational complexity, demonstrating strong potential for real-world applications.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"7 10","pages":"5482-5495"},"PeriodicalIF":0.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097425","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Exploring Latent Associations Between Exercises and Concepts With Large Language Models for Cognitive Diagnosis 用认知诊断的大语言模型探索练习和概念之间的潜在联系

IEEE transactions on artificial intelligence Pub Date : 2026-10-01 Epub Date: 2026-06-02 DOI: 10.1109/TAI.2026.3699277

Siyu Song;Haiping Ma;Changqian Wang;Xiaoshan Yu;Xingyi Zhang

{"title":"Exploring Latent Associations Between Exercises and Concepts With Large Language Models for Cognitive Diagnosis","authors":"Siyu Song;Haiping Ma;Changqian Wang;Xiaoshan Yu;Xingyi Zhang","doi":"10.1109/TAI.2026.3699277","DOIUrl":"https://doi.org/10.1109/TAI.2026.3699277","url":null,"abstract":"Cognitive diagnosis (CD) is a fundamental task in intelligent education that aims to assess students’ mastery of fine-grained knowledge concepts. Despite recent progress, most existing CD approaches heavily rely on expert-annotated Q-matrices, which are often incomplete and fail to capture latent exercise–concept associations arising in authentic problem-solving processes. To address this limitation, we propose <italic>Exploring <italic>Latent <italic>Associations between exercises and concepts with large language models (LLMs) for <italic>CD (ELACD), a novel framework that formulates latent association discovery as a dual-view cognitive inference problem. ELACD identifies latent exercise–concept associations from two complementary perspectives. From the semantic view, a large language model (LLM) enhances exercise representations by extracting implicit conceptual cues from textual content, facilitating the discovery of latent exercise–concept associations. From the structural view, we construct a student–exercise–concept interaction graph and employ graph neural networks (GNNs) to capture high-order relational dependencies that reflect empirical co-occurrence patterns in learner interactions. To effectively reconcile heterogeneous evidence from these two views, we introduce a consensus-informed fusion mechanism that adaptively integrates semantic and structural latent associations, enabling more faithful modeling of student–exercise interactions. Extensive experiments across multiple benchmark datasets demonstrate that ELACD consistently outperforms state-of-the-art CD methods. Furthermore, case studies reveal that the discovered latent associations align closely with real cognitive mastery patterns, highlighting the interpretability and practical reliability of the proposed framework.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"7 10","pages":"5722-5735"},"PeriodicalIF":0.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097445","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

EMOdiffhead: Continuously Emotional Control in Talking Head Generation via Diffusion EMOdiffhead：通过扩散持续控制说话头一代的情绪

IEEE transactions on artificial intelligence Pub Date : 2026-10-01 Epub Date: 2026-04-28 DOI: 10.1109/TAI.2026.3687506

Jian Zhang;Weijian Mai;Zhijun Zhang

{"title":"EMOdiffhead: Continuously Emotional Control in Talking Head Generation via Diffusion","authors":"Jian Zhang;Weijian Mai;Zhijun Zhang","doi":"10.1109/TAI.2026.3687506","DOIUrl":"https://doi.org/10.1109/TAI.2026.3687506","url":null,"abstract":"The task of audio-driven portrait animation involves generating a talking head video using an identity image and an audio track of speech. While many existing approaches focus on lip synchronization and video quality, few tackle the challenge of generating emotional talking head videos. The ability to control emotions is essential for producing expressive and realistic animations. To solve this problem, a novel diffusion-based emotional talking head generation method (EMOdiffhead) is presented that not only enables fine-grained control of emotion categories and intensities but also enables one-shot generation. Given the faces learned with an articulated model and expression (FLAME) 3D model’s linearity in expression modeling, the Flame-based detailed expression capture and animation (DECA) method is employed to extract expression vectors that are combined with audio to guide a diffusion model in generating videos with precise lip synchronization and rich emotional expressiveness. EMOdiffhead enables learning rich facial information from emotion-irrelevant data, thus facilitating emotional video generation. It effectively overcomes the limitations of emotional data, such as the lack of diversity in facial and background information, and addresses the absence of emotional details in emotion-irrelevant data. Extensive experiments and user studies demonstrate that the proposed method achieves state-of-the-art performance compared to other emotional portrait animation methods.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"7 10","pages":"5457-5471"},"PeriodicalIF":0.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097486","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

RGPO: Ranking-Guided Preference Optimization for Reliable Clinical Reasoning RGPO：可靠临床推理的排序导向偏好优化

IEEE transactions on artificial intelligence Pub Date : 2026-10-01 Epub Date: 2026-04-27 DOI: 10.1109/TAI.2026.3688221

Chia-Hsuan Hsu;Jun-En Ding;Hsin-Ling Hsu;Chih-Ho Hsu;Shihao Yang;Li-Hung Yao;Chun-Chieh Liao;Feng Liu;Fang-Ming Hung

{"title":"RGPO: Ranking-Guided Preference Optimization for Reliable Clinical Reasoning","authors":"Chia-Hsuan Hsu;Jun-En Ding;Hsin-Ling Hsu;Chih-Ho Hsu;Shihao Yang;Li-Hung Yao;Chun-Chieh Liao;Feng Liu;Fang-Ming Hung","doi":"10.1109/TAI.2026.3688221","DOIUrl":"https://doi.org/10.1109/TAI.2026.3688221","url":null,"abstract":"Medical question answering (QA) in healthcare systems requires advanced reasoning that integrates domain knowledge with clinical workflows. However, existing large language models (LLMs) often generate reasoning chains that lack factual accuracy and clinical reliability. We propose ranking-guided preference optimization (RGPO) for Reliable Clinical Reasoning, a framework that combines reinforcement learning with preference driven reasoning refinement to enhance clinical chain-of-thought performance in LLM-based healthcare systems. RGPO employs task-adaptive reasoning templates aligned with clinical protocols and a probabilistic evaluation mechanism that integrates with healthcare information systems, including electronic health records (EHRs) and clinical documentation, while automatically identifying and correcting low-quality reasoning chains. Unlike traditional pairwise preference methods, RGPO introduces groupwise ranking optimization based on the Bradley–Terry model with Kullback–Leibler (KL)-divergence regularization for stable clinical deployment. The framework addresses key challenges, including multimodal data integration, clinical validation, and real-world interoperability. Experiments on PubMedQA, MedQA-USMLE, and real-world validation at Far Eastern Memorial Hospital (FEMH) demonstrate consistent improvements over strong baselines. Our 2B-parameter model outperforms larger 7B–20B models, including medical-specialized variants, while maintaining computational efficiency for healthcare deployment. These results demonstrate that combining preference optimization with quality-driven refinement provides a scalable, clinically validated approach to building reliable medical LLMs for practical healthcare applications.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"7 10","pages":"5537-5550"},"PeriodicalIF":0.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149097593","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

EdgeCaps: Pruned Capsule Networks for RF Vital Sign Measurement in Real-Time EdgeCaps：用于射频生命体征实时测量的修剪胶囊网络

IEEE transactions on artificial intelligence Pub Date : 2026-10-01 Epub Date: 2026-03-26 DOI: 10.1109/TAI.2026.3696898

Prisila Ishabakaki;Michael S. Mollel;Hira Hameed;Hasan Abbas;Muhammad A. Imran;Qammer H. Abbasi

{"title":"EdgeCaps: Pruned Capsule Networks for RF Vital Sign Measurement in Real-Time","authors":"Prisila Ishabakaki;Michael S. Mollel;Hira Hameed;Hasan Abbas;Muhammad A. Imran;Qammer H. Abbasi","doi":"10.1109/TAI.2026.3696898","DOIUrl":"https://doi.org/10.1109/TAI.2026.3696898","url":null,"abstract":"Radio frequency (RF) sensing enables unobtrusive, noncontact monitoring of vital signs and has been increasingly enhanced by deep neural networks (DNNs), which excel at extracting discriminative features from complex RF signal representations. These advances have significantly improved the accuracy and robustness of RF-based health monitoring; nevertheless, state-of-the-art DNN models remain computationally intensive and unsuitable for deployment on low-cost, resource-constrained edge devices. As a result, most existing RF sensing systems offload data to high-performance platforms, leaving real-time embedded deployment underexplored. We propose <italic>EdgeCaps, a framework that compresses a high-capacity capsule network (CapsNet) into a compact student model optimized for edge deployment through structured pruning and knowledge distillation (KD). EdgeCaps transfers knowledge from a 33.7 M-parameter teacher to a 1.59M-parameter student, achieving 95.5% accuracy—within two percentage points of the teacher, while reducing parameters and computational complexity by 95%. We validate the system through continuous respiration monitoring on a Raspberry Pi 3, where the student model achieves mean inference latency below 0.3 s, RAM usage of 18%, and core temperature under <inline-formula><tex-math>$44 {}^{circ}$</tex-math></inline-formula>C, compared with the teacher’s <inline-formula><tex-math>$0.4{-}0.5$</tex-math></inline-formula> s latency, 22% RAM usage, and <inline-formula><tex-math>$51 {}^{circ}$</tex-math></inline-formula>C. These results demonstrate that EdgeCaps enables real-time, energy-efficient, and thermally stable RF-based vital sign monitoring directly on embedded devices. By combining CapsNet’s representational power with model compression and KD, this work establishes a novel pathway toward scalable, low-cost, and intelligent health monitoring systems at the edge.","PeriodicalId":73305,"journal":{"name":"IEEE transactions on artificial intelligence","volume":"7 10","pages":"5651-5663"},"PeriodicalIF":0.0,"publicationDate":"2026-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149188113","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0