检索结果-内蒙古大学图书馆

Balancing Throughput and Fair Execution of Multi-DNN Workloads on Heterogeneous Embedded Devices

IEEE Transactions on Emerging Topics in Computing 2024年 1-14页

作者： Karatzas, Andreas Anagnostopoulos, Iraklis School of Electrical Computer and Biomedical Engineering Southern Illinois University Carbondale USA

The rise of Deep Neural Networks (DNNs) has resulted in complex workloads employing multiple DNNs concurrently. This trend introduces unique challenges related to workload distribution, particularly in heterogeneous embedded systems. Current run-time managers struggle to efficiently utilize all computing components on these platforms, resulting in two major problems. First, the system throughput deteriorates due to contention on the computing resources. Second, not all DNNs are affected equally, leading to inconsistent performance levels across different models. To address these challenges, we introduce FairBoost, a framework for efficient and fair multi-DNN inference on heterogeneous embedded systems. FairBoost employs Reinforcement Learning (RL) to efficiently manage multi-DNN workloads. Additionally, it incorporates a novel numerical representation of DNN layers via a Vector Quantized Variational Auto-Encoder (VQ-VAE). Finally, it enables knowledge transfer to similar heterogeneous embedded systems without retraining and/or fine-tuning. Experimental evaluation of FairBoost over 18 DNNs and various multi-DNN scenarios shows an average throughput/fairness improvement of ×3.24. Additionally, FairBoost facilitates knowledge transfer from the initial platform, Orange Pi 5, to a new system, Odroid N2+, without any retraining or fine-tuning achieving similar gains. IEEE

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Design and usability assessment of touchscreen interface in medical devices: a case study of a neonatal incubator

引用

Research on biomedical engineering 2025年第1期41卷 1-6页

作者： Viviani, Carlos Alessandro Bassi Elias, Leonardo Abdala Calil, Saide Jorge Department of Electronics and Biomedical Engineering School of Electrical and Computer Engineering University of Campinas SP Campinas Brazil Center for Biomedical Engineering University of Campinas Campinas Brazil

来源：评论

学校读者我要写书评

暂无评论

Constraint Learning-based Optimal Power Dispatch for Active Distribution Networks with Extremely Imbalanced Data

引用

CSEE Journal of Power and Energy Systems 2024年第1期10卷 51-65页

作者： Yonghua Song Ge Chen Hongcai Zhang State Key Laboratory of Internet of Things for Smart City and Department of Electrical and Computer Engineering University of MacaoMacao 999078China School of Electrical and Computer Engineering Purdue UniversityWest LafayetteIN 47907 USA

Transition towards carbon-neutral power systems has necessitated optimization of power dispatch in active distribution networks(ADNs)to facilitate integration of distributed renewable *** to unavailability of network topology and line impedance in many distribution networks,physical model-based methods may not be applicable to their *** tackle this challenge,some studies have proposed constraint learning,which replicates physical models by training a neural network to evaluate feasibility of a decision(i.e.,whether a decision satisfies all critical constraints or not).To ensure accuracy of this trained neural network,training set should contain sufficient feasible and infeasible ***,since ADNs are mostly operated in a normal status,only very few historical samples are ***,the historical dataset is highly imbalanced,which poses a significant obstacle to neural network *** address this issue,we propose an enhanced constraint learning ***,it leverages constraint learning to train a neural network as surrogate of ADN's ***,it introduces Synthetic Minority Oversampling Technique to generate infeasible samples to mitigate imbalance of historical *** incorporating historical and synthetic samples into the training set,we can significantly improve accuracy of neural ***,we establish a trust region to constrain and thereafter enhance reliability of the *** confirm the benefits of the proposed method in achieving desirable optimality and feasibility while maintaining low computational complexity.

关键词： Deep learning demand response distribution networks imbalanced data optimal power flow

来源：评论

学校读者我要写书评

暂无评论

Discrete Current Controller for SPMSMs at Low Sampling-to-Fundamental Frequency Ratios Using Critical Damping Feedback

引用

IEEE Transactions on Power Electronics 2025年第8期40卷 11787-11798页

作者： Huang, Zhen Zhu, Rui Zhang, Yun Zanchetta, Pericle Nanchang University School of Information Engineering Nanchang330000 China Tianjin University School of Electrical and Information Engineering Tianjin300072 China University of Pavia Electrical Computer and Biomedical Engineering Department Pavia27100 Italy

The discrete current controllers can generally ensure high-performance machine control. However, for high-speed machines associated with low sampling-to-fundamental frequency ratios (SFR), the unavoidable disturbances will seriously affect the control performance. To address this issue, an advanced discrete current controller is proposed in this article. The critical damping feedback is used to achieve the maximum damping ratio of the disturbance loop at any SFR. Compared with the existing discrete controllers, the proposed one can significantly reduce the impact of multiple parameter mismatches on the performance of the controller while improving the current tracking response and enhancing the antidisturbance capability. Finally, the experimental results on a 10 krpm high-speed permanent magnet synchronous machine with the lowest SFR of 8 have proven the feasibility and advantages of the addressed current controller. © 1986-2012 IEEE.

关键词： Damping

来源：评论

学校读者我要写书评

暂无评论

Mamba State Space Model (SSM) Method for the Automatic Segmentation of Organs at Risk from Head and Neck CT Images 4

Mamba State Space Model (SSM) Method for the Automatic Segme...

引用

4th International Conference on electrical, computer, Communications and Mechatronics engineering, ICECCME 2024

作者： Vagenas, Theodoros P. Georgas, Konstantinos Matsopoulos, George K. School of Electrical and Computer Engineering National Technical University of Athens Biomedical Engineering Laboratory Athens Greece

ISBN: (纸本)9798350391183

Radiotherapy (RT) is one of the main treatment modalities using high-energy radiation to eliminate cancer cells in Head and Neck cancer (HNC) for patients worldwide. Advances such as intensity-modulated radiotherapy (IMRT) improve the precision of targeting tumors and prevent damaging healthy tissues. Segmentation of Organs At Risk (OARs) on Computed Tomography (CT) images is crucial for RT treatment planning in order to avoid damaging healthy structures during the sessions. However, manual delineation of OARs is challenging as their size varies and both small and large organs are present in images with low contrast against the background. In this regard, manual delineation cannot be performed in a timely manner and suffers from potential interobserver variability among clinicians, leading to delays in clinical workflow. In this study, we propose an automatic Deep Learning segmentation framework utilizing effectively the recently introduced Mamba State Space Models (SSM) to localize and delineate multiple OARs with high accuracy from the 3D HNC CT images. Due to the Mamba-based encoder, the model can learn the long-range dependencies of the largest organs by efficiently handling large sequences while also segmenting the smaller ones. The proposed model was evaluated on two public datasets, demonstrating superior performance with an average Dice Similarity Coefficient (DSC) of 75.45% and a Hausdorff distance of 8.22 mm against established deep learning models, such as UNet-based and Transformer-based, in segmenting 9 OARs, indicating promising results to support the clinical workflow. © 2024 IEEE.

关键词： computerized tomography

来源：评论

学校读者我要写书评

暂无评论

Cavity-agnostic acoustofluidic manipulations enabled by guided flexural waves on a membrane acoustic waveguide actuator

引用

Microsystems & Nanoengineering 2024年第2期10卷 153-171页

作者： Philippe Vachon Srinivas Merugu Jaibir Sharma Amit Lal Eldwin J.Ng Yul Koh Joshua E.-Y.Lee Chengkuo Lee Department of Electrical and Computer Engineering National University of SingaporeSingaporeSingapore Institute of Microelectronics Agency for ScienceTechnology and Research(A*STAR)SingaporeSingapore SonicMEMS Laboratory School of Electrical and Computer EngineeringCornell UniversityIthacaNYUSA School of Electrical and Data Engineering University of Technology SydneyUltimoNSWAustralia

This article presents an in-depth exploration of the acoustofluidic capabilities of guided flexural waves(GFWs)generated by a membrane acoustic waveguide actuator(MAWA).By harnessing the potential of GFWs,cavity-agnostic advanced particle manipulation functions are achieved,unlocking new avenues for microfluidic systems and lab-on-a-chip *** localized acoustofluidic effects of GFWs arising from the evanescent nature of the acoustic fields they induce inside a liquid medium are numerically investigated to highlight their unique and promising *** traditional acoustofluidic technologies,the GFWs propagating on the MAWA’s membrane waveguide allow for cavity-agnostic particle manipulation,irrespective of the resonant properties of the fluidic ***,the acoustofluidic functions enabled by the device depend on the flexural mode populating the active region of the membrane *** demonstrations using two types of particles include in-sessile-droplet particle transport,mixing,and spatial separation based on particle diameter,along with streaming-induced counter-flow virtual channel generation in microfluidic PDMS *** experiments emphasize the versatility and potential applications of the MAWA as a microfluidic platform targeted at lab-on-a-chip development and showcase the MAWA’s compatibility with existing microfluidic systems.

关键词： waveguide fluid cavity

来源：评论

学校读者我要写书评

暂无评论

MMInstruct: a high-quality multi-modal instruction tuning dataset with extensive diversity

引用

Science China(Information Sciences) 2024年第12期67卷 36-51页

作者： Yangzhou LIU Yue CAO Zhangwei GAO Weiyun WANG Zhe CHEN Wenhai WANG Hao TIAN Lewei LU Xizhou ZHU Tong LU Yu QIAO Jifeng DAI School of Computer Science Nanjing University School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University Shanghai AI Laboratory School of Computer Science Fudan University Department of Information Engineering The Chinese University of Hong Kong SenseTime Research Department of Electronic Engineering Tsinghua University

Despite the effectiveness of vision-language supervised fine-tuning in enhancing the performance of vision large language models(VLLMs), existing visual instruction tuning datasets include the following limitations.(1) Instruction annotation quality: despite existing VLLMs exhibiting strong performance,instructions generated by those advanced VLLMs may still suffer from inaccuracies, such as hallucinations.(2) Instructions and image diversity: the limited range of instruction types and the lack of diversity in image data may impact the model's ability to generate diversified and closer to real-world scenarios outputs. To address these challenges, we construct a high-quality, diverse visual instruction tuning dataset MMInstruct,which consists of 973k instructions from 24 domains. There are four instruction types: judgment, multiplechoice, long visual question answering, and short visual question answering. To construct MMInstruct, we propose an instruction generation data engine that leverages GPT-4V, GPT-3.5, and manual correction. Our instruction generation engine enables semi-automatic, low-cost, and multi-domain instruction generation at 1/6 the cost of manual construction. Through extensive experiment validation and ablation experiments,we demonstrate that MMInstruct could significantly improve the performance of VLLMs, e.g., the model fine-tuning on MMInstruct achieves new state-of-the-art performance on 10 out of 12 benchmarks. The code and data shall be available at https://***/yuecao0119/MMInstruct.

关键词： instruction tuning multi-modal multi-domain dataset vision large language model

来源：评论

学校读者我要写书评

暂无评论

EEG Prediction with Deep Recurrent Learning

EEG Prediction with Deep Recurrent Learning

引用

2024 IEEE International Conference on Consumer Electronics, ICCE 2024

作者： Zou, Jiadao Zhang, Qingxue Purdue Univeristy School of Engineering and Technology Electrical and Computer Engineering IN United States Purdue Univeristy School of Engineering and Technology Electrical and Computer Engineering Biomedical Engineering IN United States

ISBN: (纸本)9798350324136

Electroencephalogram (EEG) reveals the brain dynamics that are of a great value in clinical applications, e.g., epilepsy diagnosis, cognitive status, and other disorders. Intracranial Electroencephalogram (iEEG) is encoding rich dynamics, with its invasive electrodes that go deep into the brain. Despite the high precision, the iEEG surgery may limit its use in daily lives due to the risk, setup time and cost. In this study, we propose a deep learning framework that can reconstruct iEEG from scalp EEG, thereby demonstrating the potential of convenient deep brain monitoring in daily lives. We leverage a long short-Term memory (LSTM) model for ultra-deep iEEG prediction from the non-invasive scalp EEG. Our simulation has demonstrated that, our deep learning model, with the nonlinear dynamics mining capability, facilitates the deep source signal estimation performance, compared with the linear model. Furthermore, we find that the estimation performance may be improved with selection of more EEG measurement sites. This study will greatly advance brain dynamics decoding and broadly benefit diverse brain-computer-interface and smart health applications. © 2024 IEEE.

关键词： Brain computer Interface Electroencephalogram Recurrent Learning Smart Health

来源：评论

学校读者我要写书评

暂无评论

Ultrasonically actuated neural probes for reduced trauma and inflammation in mouse brain

引用

Microsystems & Nanoengineering 2022年第6期8卷 137-151页

作者： Po-Cheng Chen Catharine G.Young Chris B.Schaffer Amit Lal SonicMEMS Laboratory School of Electrical and Computer EngineeringCornell UniversityIthacaNYUSA Meinig School of Biomedical Engineering Cornell UniversityIthacaNYUSA

electrical neural recordings measured using direct electrical interfaces with neural tissue suffer from a short lifespan because the signal strength decreases over *** inflammatory response to the inserted microprobe can create insulating tissue over the electrical interfaces,reducing the recorded signal below noise *** of the factors contributing to this inflammatory response is the tissue damage caused during probe ***,we explore the use of ultrasonic actuation of the neural probe during insertion to minimize tissue damage in *** neural microprobes were designed and fabricated with integrated electrical recording sites and piezoelectric *** microprobes were actuated at ultrasonic frequencies using integrated piezoelectric *** microprobes were inserted into mouse brains under a glass window over the brain surface to image the tissue surrounding the probe using two-photon *** mechanical force required to penetrate the tissue was reduced by a factor of 2–3 when the microprobe was driven at ultrasonic *** histology at the probe insertion site showed a reduced area of damage and decreased microglia counts with increasing ultrasonic actuation of the *** imaging of the microprobe over weeks demonstrated stabilization of the inflammatory *** of electrical signals from neurons over time suggests that microprobes inserted using ultrasound have a higher signal-tonoise ratio over an extended time period.

关键词： microprobe insertion electrical

来源：评论

学校读者我要写书评

暂无评论

Predefined-time distributed multiobjective optimization for network resource allocation

引用

Science China(Information Sciences) 2023年第7期66卷 58-72页

作者： Kunpeng ZHANG Lei XU Xinlei YI Zhengtao DING Karl H.JOHANSSON Tianyou CHAI Tao YANG State Key Laboratory of Synthetical Automation for Process Industries Northeastern University School of Electrical Engineering and Computer Science KTH Royal Institute of Technology School of Electrical and Electronic Engineering The University of Manchester

We consider the multiobjective optimization problem for the resource allocation of the multiagent network, where each agent contains multiple conflicting local objective functions. The goal is to find compromise solutions minimizing all local objective functions subject to resource constraints as much as possible, i.e., the Pareto optimums. To this end, we first reformulate the multiobjective optimization problem into one single-objective distributed optimization problem by using the weighted Lppreference index,where the weighting factors of all local objective functions are obtained from the optimization procedure so that the optimizer of the latter is the desired Pareto optimum of the former. Next, we propose novel predefined-time algorithms to solve the reformulated problem by time-based generators. We show that the reformulated problem is solved within a predefined time if the local objective functions are strongly convex and smooth. Moreover, the settling time can be arbitrarily preset since it does not depend on the initial values and designed parameters. Finally, numerical simulations are presented to illustrate the effectiveness of the proposed algorithms.

关键词： distributed optimization multiobjective optimization predefined-time algorithms time-based generators weighted ${L_p}$ preference index

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：