检索结果-内蒙古大学图书馆

GACL-Net:Hybrid Deep Learning Framework for Accurate Motor Imagery Classification in Stroke Rehabilitation

computers, Materials & Continua 2025年第4期83卷 517-536页

作者： Chayut Bunterngchit Laith H.Baniata Mohammad H.Baniata Ashraf ALDabbas Mohannad A.Khair Thanaphon Chearanai Sangwoo Kang State Key Laboratory of Multimodal Artificial Intelligence Systems Institute of AutomationChinese Academy of SciencesBeijing100190China School of Computing Gachon UniversitySeongnam13120Republic of Korea Computer Science Department Faculty of Information TechnologyThe World Islamic Sciences and Education UniversityAmman11947Jordan Intelligent Systems Department Faculty of Artificial IntelligenceAl-Balqa Applied UniversityAl-Salt19117Jordan IT Infrastructure Department Qatrana Cement CompanyAmman11821Jordan Division of Industrial and Logistics Engineering Technology Faculty of Engineering and TechnologyKing Mongkut’s University of Technology North BangkokRayong CampusRayong21120Thailand

Stroke is a leading cause of death and disability worldwide,significantly impairing motor and cognitive *** rehabilitation is often hindered by the heterogeneity of stroke lesions,variability in recovery patterns,and the complexity of electroencephalography(EEG)signals,which are often contaminated by *** classification of motor imagery(MI)tasks,involving the mental simulation of movements,is crucial for assessing rehabilitation strategies but is challenged by overlapping neural signatures and patient-specific *** address these challenges,this study introduces a graph-attentive convolutional long short-term memory(LSTM)network(GACL-Net),a novel hybrid deep learning model designed to improve MI classification accuracy and ***-Net incorporates multi-scale convolutional blocks for spatial feature extraction,attention fusion layers for adaptive feature prioritization,graph convolutional layers to model inter-channel dependencies,and bidi-rectional LSTM layers with attention to capture temporal *** on an open-source EEG dataset of 50 acute stroke patients performing left and right MI tasks,GACL-Net achieved 99.52%classification accuracy and 97.43%generalization accuracy under leave-one-subject-out cross-validation,outperforming existing state-of-the-art ***,its real-time processing capability,with prediction times of 33–56 ms on a T4 GPU,underscores its clinical potential for real-time neurofeedback and adaptive *** findings highlight the model’s potential for clinical applications in assessing rehabilitation effectiveness and optimizing therapy plans through precise MI classification.

关键词： Motor imagery EEG stroke rehabilitation deep learning brain-computer interface

来源：评论

学校读者我要写书评

暂无评论

Boolean matrix compressed sensing

Boolean matrix compressed sensing

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Liu, Qiang Soleymani, Mahdi Mahdavifar, Hessam National Laboratory of Industrial Control Technology Institute of Cyber-Systems and Control Zhejiang University Hangzhou310027 China Halıcıoğlu Data Science Institute University of California La Jolla San DiegoCA92093 United States Department of Electrical and Computer Engineering Northeastern University BostonMA02115 United States

ISBN: (纸本)9798350368741

In real-world datasets, leveraging the low-rank and sparsity properties enables developing efficient algorithms across a diverse array of data-related tasks, including compression, compressed sensing, matrix completion, etc. Notably, these two properties often coexist in certain real-world datasets, especially in Boolean datasets and quantized real-valued datasets. To harness the advantages of low-rank and sparsity simultaneously, we adopt a technique inspired by compressed sensing and Boolean matrix completion. Our approach entails compressing a low-rank sparse Boolean matrix by performing inner product operations with a randomly generated Boolean matrix. We then propose a decoding algorithms based on message-passing techniques to recover the original matrix. Our experiments demonstrate superior recovery performance of our proposed algorithms compared to Boolean matrix completion, with equal measurement requirements. © 2025 IEEE.

关键词： Belief Propagation Boolean Matrix Compressed Sensing Matrix Completion

来源：评论

学校读者我要写书评

暂无评论

Competitive Multi-armed Bandit Games for Resource Sharing

引用

IEEE Transactions on Mobile Computing 2025年

作者： Li, Hongbo Duan, Lingjie The Ohio State University Department of Electrical and Computer Engineering ColumbusOH43210 United States Singapore University of Technology and Design Pillar of Engineering Systems and Design 487372 Singapore

In modern resource-sharing systems, multiple agents access limited resources with unknown stochastic conditions to perform tasks. When multiple agents access the same resource (arm) simultaneously, they compete for successful usage, leading to contention and reduced rewards. This motivates our theoretical study of competitive multi-armed bandit (CMAB) games. In this paper, we study a new N-player K-arm competitive MAB game, where non-myopic players (agents) compete with each other to form diverse private estimations of unknown arms over time. Their possible collisions on the same arms and the time-varying nature of arm rewards make the policy analysis here more involved than the existing studies for myopic players. We explicitly analyze the threshold-based structures of the social optimum and the existing selfish policy, showing that the latter causes prolonged convergence times (Formula presented), while the socially optimal policy with coordinated communication reduces it to (Formula presented). Based on the policy comparison, we prove that the competition among selfish players for the best arm can result in an infinite price of anarchy (PoA), indicating an arbitrarily large efficiency loss compared to the social optimum. We further prove that no informational (non-monetary) mechanism (including Bayesian persuasion) can reduce the infinite PoA, as strategic misreporting by non-myopic players undermines such approaches. To address this, we propose a Combined Informational and Side-Payment (CISP) mechanism, which provides socially optimal arm recommendations with proper informational and monetary incentives to players according to their diverse and time-varying private beliefs. Our CISP mechanism keeps ex-post budget balanced for the social planner and ensures truthful reporting from players, thereby achieving the minimum PoA = 1 and the same convergence time as the social optimum. © 2002-2012 IEEE.

关键词： Budget control

来源：评论

学校读者我要写书评

暂无评论

RoSTE: An Efficient Quantization-Aware Supervised Fine-Tuning Approach for Large Language Models

arXiv

引用

arXiv 2025年

作者： Wei, Quan Yau, Chung-Yiu Wai, Hoi-To Zhao, Yang Kang, Dongyeop Park, Youngsuk Hong, Mingyi Department of Electrical and Computer Engineering University of Minnesota United States Department of Systems Engineering and Engineering Management The Chinese University of Hong Kong Hong Kong Department of Computer Science and Engineering University of Minnesota United States Amazon Web Services United States

Supervised fine-tuning is a standard method for adapting pre-trained large language models (LLMs) to downstream tasks. Quantization has been recently studied as a post-training technique for efficient LLM deployment. To obtain quantized fine-tuned LLMs, conventional pipelines would first fine-tune the pre-trained models, followed by post-training quantization. This often yields suboptimal performance as it fails to leverage the synergy between fine-tuning and quantization. To effectively realize low-bit quantization of weights, activations, and KV caches in LLMs, we propose an algorithm named Rotated Straight-Through-Estimator (RoSTE), which combines quantization-aware supervised fine-tuning (QA-SFT) with an adaptive rotation strategy that identifies an effective rotation configuration to reduce activation outliers. We provide theoretical insights on RoSTE by analyzing its prediction error when applied to an overparameterized least square quantized training problem. Our findings reveal that the prediction error is directly proportional to the quantization error of the converged weights, which can be effectively managed through an optimized rotation configuration. Experiments on Pythia, Qwen and Llama models of different sizes demonstrate the effectiveness of RoSTE. Compared to existing post-SFT quantization baselines, our method consistently achieves superior performances across various tasks and different LLM architectures. Our code is available at https://***/Hong-Lab-UMN-ECE/RoSTE. © 2025, CC BY.

关键词： Digital elevation model

来源：评论

学校读者我要写书评

暂无评论

Accelerating OTA Circuit Design: Transistor Sizing Based on a Transformer Model and Precomputed Lookup Tables

arXiv

引用

arXiv 2025年

作者： Ghosh, Subhadip Gebru, Endalk Y. Kashyap, Chandramouli V. Harjani, Ramesh Sapatnekar, Sachin S. Department of Electrical and Computer Engineering University of Minnesota MinneapolisMN United States Cadence Design Systems PortlandOR United States

Device sizing is crucial for meeting performance specifications in operational transconductance amplifiers (OTAs), and this work proposes an automated sizing framework based on a transformer model. The approach first leverages the driving-point signal flow graph (DPSFG) to map an OTA circuit and its specifications into transformer-friendly sequential data. A specialized tokenization approach is applied to the sequential data to expedite the training of the transformer on a diverse range of OTA topologies, under multiple specifications. Under specific performance constraints, the trained transformer model is used to accurately predict DP-SFG parameters in the inference phase. The predicted DP-SFG parameters are then translated to transistor sizes using a precomputed look-up table-based approach inspired by the gm/Id methodology. In contrast to previous conventional or machine-learning-based methods, the proposed framework achieves significant improvements in both speed and computational efficiency by reducing the need for expensive SPICE simulations within the optimization loop;instead, almost all SPICE simulations are confined to the onetime training phase. The method is validated on a variety of unseen specifications, and the sizing solution demonstrates over 90% success in meeting specifications with just one SPICE simulation for validation, and 100% success with 3–5 additional SPICE simulations. © 2025, CC BY.

关键词： Table lookup

来源：评论

学校读者我要写书评

暂无评论

Extraction of Common Mode Parasitic Capacitance in Balance Filter for The Prediction of EMI Noise Suppression

Extraction of Common Mode Parasitic Capacitance in Balance F...

引用

Annual IEEE Conference on Applied Power Electronics Conference and Exposition (APEC)

作者： Qiuzhe Yang Xingyu Chen Zijian Wang Qiang Li Bradley Department of Electrical and Computer Engineering Center for Power Electronics Systems Virginia Polytechnic and State University Blacksburg VA USA

ISBN: (数字)9798331516116

ISBN: (纸本)9798331516123

This paper proposes a new method to extract the common mode parasitic capacitance in the balance network for the totem-pole converter. Filter concept is applied to the balance network to predict the EMI noise with the extracted parasitic capacitance. With the proposed method, the EMI noise prediction matches the experiment very well within 30MHz.

关键词： Matched filters Voltage measurement Noise Electromagnetic interference Noise reduction Switches Predictive models Power electronics Noise measurement Parasitic capacitance

来源：评论

学校读者我要写书评

暂无评论

Prescriptive Analytics for Logistics: Bridging Predictive Models and Actionable Decisions Through Advanced Optimization Techniques

Prescriptive Analytics for Logistics: Bridging Predictive Mo...

引用

Innovative Research in Applied Science, engineering and Technology (IRASET),Conference on

作者： Oussama El Azzouzy Tarik Chanyour Said Jai Andaloussi Department of Mathematics and Computer Science Laboratory of Data Engineering and Intelligent Systems(LIDSI) Hassan II University Casablanca Morocco

ISBN: (数字)9798331532970

ISBN: (纸本)9798331532987

Modern logistics is facing increasing challenges due to the growing complexity of supply chains, high consumer expectations, and economic and environmental pressures. While predictive analytics, powered by artificial intelligence (AI) and massive data, has transformed the ability to forecast trends and risks, its application remains limited by the lack of effective mechanisms to convert these predictions into actionable decisions in dynamic and constrained environments. This paper proposes an innovative hybrid approach that combines advanced predictive models, notably the Temporal Fusion Transformer (TFT), with prescriptive algorithms, such as linear programming and deep reinforcement learning (DRL). The TFT demonstrated superior performance in terms of predictive accuracy (significant reduction in MAE, RMSE, and MAPE), while the prescriptive algorithms optimized complex logistics decisions in real time, even in uncertain environments. This integrated methodology was validated through use cases simulating critical logistics scenarios, such as delivery route optimization. Experimental results show a substantial improvement in operational efficiency, with reductions in costs, lead times and resource utilization. The prescriptive approach made it possible to automate complex decisions, while taking into account multiple constraints, such as costs, lead times and sustainability objectives. By integrating real-time data collected via IoT devices and logistics management systems, this architecture offers a robust, adaptable and scalable solution. It thus meets the requirements of modern logistics by enabling proactive, optimized and resilient decision-making. This research makes a significant contribution by bridging the gap between predictive and prescriptive analysis, while laying the foundations for new applications in the field of intelligent logistics.

关键词： Costs Supply chains Predictive models Prediction algorithms Linear programming Transformers Real-time systems Thin film transistors Resource management Optimization

来源：评论

学校读者我要写书评

暂无评论

A Wearable Paper-Based Hybrid Energy Harvester from Human Sweat and Ambient Moisture 38

A Wearable Paper-Based Hybrid Energy Harvester from Human Sw...

引用

38th IEEE International Conference on Micro Electro Mechanical systems, MEMS 2025

作者： Gao, Yang Choi, Seokheun State University of New York at Binghamton Bioelectronics & Microsystems Laboratory Department of Electrical & Computer Engineering BinghamtonNY United States State University of New York at Binghamton Center for Research in Advanced Sensing Technologies and Environmental Sustainability BinghamtonNY United States

ISBN: (纸本)9798331508890

We present a wearable, paper-based energy harvester that combines a sweat-driven microbial fuel cell (MFC) with an ambient moisture-enabled electric generator (MEG), capable of simultaneously or independently harnessing energy from both human sweat and the environment. Paper stands out as an ideal substrate for wearable applications, offering flexibility and easy integration through straightforward printing methods. Due to the simple design of the MFC and MEG, both devices can be seamlessly integrated into a single sheet of paper. The biocatalyst-embedded anode of the MFC is positioned directly against the skin for optimal sweat absorption, while the hygroscopic region of the MEG faces outward to efficiently capture ambient moisture. At low humidity levels, the energy harvesting is primarily driven by the MFC, while in the absence of sweat, the MEG takes over as the main power source. This dual functionality ensures continuous and sustainable power generation, making the system highly efficient and reliable for long-term wearable applications. © 2025 IEEE.

关键词： Microbial fuel cells

来源：评论

学校读者我要写书评

暂无评论

Observation of merging mechanical bound states in the continuum in phononic crystals 15

Observation of merging mechanical bound states in the contin...

引用

Photonic and Phononic Properties of Engineered Nanostructures XV 2025

作者： Liu, Shengyan Tong, Hao Fang, Kejie Holonyak Micro and Nanotechnology Laboratory Department of Electrical and Computer Engineering University of Illinois Urbana-Champaign UrbanaIL61801 United States Illinois Quantum Information Science and Technology Center University of Illinois Urbana-Champaign UrbanaIL61801 United States

ISBN: (纸本)9781510685024

Mechanical bound states in the continuum (BICs) have gradually become another popular method, aside from traditional band engineering, for obtaining mechanical resonators with high frequency and high quality factors. In our previous work, we demonstrated symmetry-protected mechanical BICs and coupled them with coexisting optical band-edge modes to achieve release-free optomechanical crystals. However, another important type of BIC — accidental BIC has never been realized in phononic systems. Here, we observe accidental BICs in a high-aspect-ratio gallium arsenide phononic crystal grating, their merging process with symmetry-protected BICs, and the subsequent enhancement of mechanical quality factors. This discovery provides a potential method for enhancing the performance of BIC based optomechanical crystals and can also be applied in fields such as sensing and quantum transduction. © 2025 SPIE.

关键词： Phonons

来源：评论

学校读者我要写书评

暂无评论

Bio-Inspired Home Localization Using Event-Based Vision and Spiking Neural Networks in Simulated Environment

Bio-Inspired Home Localization Using Event-Based Vision and ...

引用

IEEE International Conference on Simulation, Modeling, and Programming for Autonomous Robots (SIMPAR)

作者： Rizwana Kausar Fakhreddine Zayer Vidya Sudevan Jaime Viegas Jorge Dias Center for Autonomous Robotic Systems Khalifa University UAE Department of Electrical Engineering Khalifa University UAE Department of Computer and Information Engineering Khalifa University UAE

ISBN: (数字)9798331516857

ISBN: (纸本)9798331516864

In this work, we present a bio-inspired approach for home localization using event-based visual data and spiking convolutional neural networks (S-CNNs) in a simulated environment within NVIDIA Omniverse. Drawing inspiration from the navigational strategies of insects, which utilize visual cues and homing vectors to return to their nests, our approach leverages event cameras that mimic the asynchronous change-driven visual processing of insect eyes. To process this sparse, event-driven input, we employ spiking neural networks (SNNs), which mirror the spike-based transmission of information in biological neural systems. The proposed system utilizes a quadrotor equipped with an event camera to capture dynamic, asynchronous visual data and processes it using an S-CNN trained to estimate relative home vectors. This vector encoding method is inspired by the unit-circle representation of gaze directions observed in insect homing behavior, as outlined in recent studies. By integrating event-based vision and SNNs, our approach ensures energy-efficient computation, robustness to lighting variations, and adaptability to dynamic scenes. We validate the framework in a 3D environment modeled within ISAAC Sim, where the quadrotor autonomously navigates back to a designated "nest" location. Comparative analysis with conventional frame-based neural networks demonstrates the superiority of the proposed system in terms of efficiency, robustness, and accuracy. This work establishes a novel bio-inspired framework for integrating event-based data and spiking neural networks, paving the way for energy-efficient localization in robotics and smart home environments. Future work will explore the deployment of the system in a multi-quadrotor environment to coordinate collaborative tasks including the integration of multi-modal sensory inputs such as Camera, IMU, Gas sensors and GPS data, and the extension of the framework to real-world settings for further validation and scalability.

关键词： Location awareness Visualization Adaptation models Event detection Insects Computational modeling Spiking neural networks Vectors Robustness Quadrotors

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：