检索结果-内蒙古大学图书馆

17th IEEE International Symposium on Embedded Multicore/Many-core Systems-on-Chip, MCSoC 2024

作者： Aliyev, Ilkin Svoboda, Kama Adegbija, Tosiron Fellous, Jean-Marc University of Arizona Department of Electrical & Computer Engineering TucsonAZ United States University of Arizona Departments of Psychology and Biomedical Engineering TucsonAZ United States

ISBN: (纸本)9798331530471

Spiking Neural Networks (SNNs) are inspired by the sparse and event-driven nature of biological neural processing, and offer the potential for ultra-low-power artificial intelligence. However, realizing their efficiency benefits requires specialized hardware and a co-design approach that effectively leverages sparsity. We explore the hardware-software co-design of sparse SNNs, examining how sparsity representation, hardware architectures, and training techniques influence hardware efficiency. We analyze the impact of static and dynamic sparsity, discuss the implications of different neuron models and encoding schemes, and investigate the need for adaptability in hardware designs. Our work aims to illuminate the path towards embedded neuro-morphic systems that fully exploit the computational advantages of sparse SNNs. © 2024 IEEE.

关键词： Neural network models

来源：评论

学校读者我要写书评

暂无评论

Exploring the trade-off between deep-learning and explainable models for brain-machine interfaces 38

Exploring the trade-off between deep-learning and explainabl...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Cubillos, Luis H. Revach, Guy Mender, Matthew J. Costello, Joseph T. Temmar, Hisham Hite, Aren Zutshi, Diksha Wallace, Dylan M. Ni, Xiaoyong Kelberman, Madison M. Willsey, Matthew S. van Sloun, Ruud J.G. Shlezinger, Nir Patil, Parag Draelos, Anne Chestek, Cynthia A. Departments of Electrical & Computer Engineering Biomedical Engineering Robotics Computational Medicine & Bioinformatics and Neurosurgery University of Michigan United States Biointerfaces Institute Neuroscience Institute University of Michigan United States Department of Information Technology and Electrical Engineering ETH Zürich Switzerland Department of Electrical Engineering Eindhoven University of Technology Netherlands School of Electrical and Computer Engineering Ben-Gurion University Israel

People with brain or spinal cord-related paralysis often need to rely on others for basic tasks, limiting their independence. A potential solution is brain-machine interfaces (BMIs), which could allow them to voluntarily control external devices (e.g., robotic arm) by decoding brain activity to movement commands. In the past decade, deep-learning decoders have achieved state-of-the-art results in most BMI applications, ranging from speech production to finger control. However, the'black-box' nature of deep-learning decoders could lead to unexpected behaviors, resulting in major safety concerns in real-world physical control scenarios. In these applications, explainable but lower-performing decoders, such as the Kalman filter (KF), remain the norm. In this study, we designed a BMI decoder based on KalmanNet, an extension of the KF that augments its operation with recurrent neural networks to compute the Kalman gain. This results in a varying "trust" that shifts between inputs and dynamics. We used this algorithm to predict finger movements from the brain activity of two monkeys. We compared KalmanNet results offline (pre-recorded data, n = 13 days) and online (real-time predictions, n = 5 days) with a simple KF and two recent deep-learning algorithms: tcFNN (non-ReFIT version) and LSTM. KalmanNet achieved comparable or better results than other deep learning models in offline and online modes, relying on the dynamical model for stopping while depending more on neural inputs for initiating movements. We further validated this mechanism by implementing a heteroscedastic KF that used the same strategy, and it also approached state-of-the-art performance while remaining in the explainable domain of standard KFs. However, we also see two downsides to KalmanNet. KalmanNet shares the limited generalization ability of existing deep-learning decoders, and its usage of the KF as an inductive bias limits its performance in the presence of unseen noise distributions. Despite thi

关键词：

来源：评论

学校读者我要写书评

暂无评论

Sparsity-Aware Hardware-Software Co-Design of Spiking Neural Networks: An Overview

arXiv

引用

arXiv 2024年

作者： Aliyev, Ilkin Svoboda, Kama Adegbija, Tosiron Fellous, Jean-Marc Department of Electrical & Computer Engineering Departments of Psychology and Biomedical Engineering University of Arizona TucsonAZ United States

Spiking Neural Networks (SNNs) are inspired by the sparse and event-driven nature of biological neural processing, and offer the potential for ultra-low-power artificial intelligence. However, realizing their efficiency benefits requires specialized hardware and a co-design approach that effectively leverages sparsity. We explore the hardware-software co-design of sparse SNNs, examining how sparsity representation, hardware architectures, and training techniques influence hardware efficiency. We analyze the impact of static and dynamic sparsity, discuss the implications of different neuron models and encoding schemes, and investigate the need for adaptability in hardware designs. Our work aims to illuminate the path towards embedded neuromorphic systems that fully exploit the computational advantages of sparse SNNs. © 2024, CC BY.

关键词： Neural network models

来源：评论

学校读者我要写书评

暂无评论

Quantifying Strain Dependence of Multi-Frequency Shearwave Elasticity Imaging

Quantifying Strain Dependence of Multi-Frequency Shearwave E...

引用

2023 IEEE Western New York Image and Signal Processing Workshop, WNYISPW 2023

作者： Khan, Siladitya Goswami, Soumya Feng, Fan Doyley, Marvin M. McAleavey, Stephen A. University of Rochester Departments of Biomedical and Computer Engineering RochesterNY14627 United States University of Rochester Rochester Center for Biomedical Ultrasound RochesterNY14627 United States University of Rochester Goergen Institute for Data Science RochesterNY14627 United States University of Rochester Departments of Electrical and Computer Engineering RochesterNY14627 United States

ISBN: (纸本)9798350329698

Nonlinear shear modulus (NLSM) can potentially differentiate benign and malignant pathologies. Previous studies demonstrated the discriminatory value of mapping quantitative estimates of material non-linearity, yet estimates remain biased due to unaccounted shear-wave dispersion within the tissue. We demonstrate a finite element simulation of an inversion technique that combines traditional quasi-static and dynamic elasticity imaging to generate a multi-frequency estimate of the third-order shear modulus A(x,z,ω) over a broad bandwidth (200-600 Hz). We show that our reconstruction results are superior to those obtained by traditional group speed by a factor of two in terms of contrast enhancement measures. Experimental ultrasound elastography on a custom oil-in-gelatin emulsion confirms the presence of dispersion in the NLSM parameter. Constitutive hyper-viscoelastic model responses of spectral NLSM are simulated based on matched linear shear modulus data of experimental phantoms. Hyper-elastic Ogden with a Voigt loss is shown to describe NLSM behavior at frequencies greater than 400 Hz. Reconstruction accuracy in the presence of simulated tracking and correlation noise demonstrates the robustness of our estimates. © 2023 IEEE.

关键词： Elastic moduli

来源：评论

学校读者我要写书评

暂无评论

Exploring the trade-off between deep-learning and explainable models for brain-machine interfaces 24

Exploring the trade-off between deep-learning and explainabl...

引用

Proceedings of the 38th International Conference on Neural Information Processing Systems

作者： Luis H. Cubillos Guy Revach Matthew J. Mender Joseph T. Costello Hisham Temmar Aren Hite Diksha Zutshi Dylan M. Wallace Xiaoyong Ni Madison M. Kelberman Matthew S. Willsey Ruud J.G. van Sloun Nir Shlezinger Parag Patil Anne Draelos Cynthia A. Chestek Departments of Electrical & Computer Engineering Biomedical Engineering Robotics Computational Medicine & Bioinformatics and Neurosurgery University of Michigan Department of Information Technology and Electrical Engineering ETH Zürich Switzerland Departments of Electrical & Computer Engineering Biomedical Engineering Robotics Computational Medicine & Bioinformatics and Neurosurgery University of Michigan and Biointerfaces Institute and Neuroscience Institute University of Michigan Department of Electrical Engineering Eindhoven University of Technology Netherlands School of Electrical and Computer Engineering Ben-Gurion University Israel

ISBN: (纸本)9798331314385

People with brain or spinal cord-related paralysis often need to rely on others for basic tasks, limiting their independence. A potential solution is brain-machine interfaces (BMIs), which could allow them to voluntarily control external devices (e.g., robotic arm) by decoding brain activity to movement commands. In the past decade, deep-learning decoders have achieved state-of-the-art results in most BMI applications, ranging from speech production to finger control. However, the 'black-box' nature of deep-learning decoders could lead to unexpected behaviors, resulting in major safety concerns in real-world physical control scenarios. In these applications, explainable but lower-performing decoders, such as the Kalman filter (KF), remain the norm. In this study, we designed a BMI decoder based on Kalman-Net, an extension of the KF that augments its operation with recurrent neural networks to compute the Kalman gain. This results in a varying "trust" that shifts between inputs and dynamics. We used this algorithm to predict finger movements from the brain activity of two monkeys. We compared KalmanNet results offline (pre-recorded data, n =13 days) and online (real-time predictions, n = 5 days) with a simple KF and two recent deep-learning algorithms: tcFNN (non-ReFIT version) and LSTM. KalmanNet achieved comparable or better results than other deep learning models in offline and online modes, relying on the dynamical model for stopping while depending more on neural inputs for initiating movements. We further validated this mechanism by implementing a heteroscedastic KF that used the same strategy, and it also approached state-of-the-art performance while remaining in the explainable domain of standard KFs. However, we also see two downsides to KalmanNet. KalmanNet shares the limited generalization ability of existing deep-learning decoders, and its usage of the KF as an inductive bias limits its performance in the presence of unseen noise distributions. Despite th

关键词：

来源：评论

学校读者我要写书评

暂无评论

Spatial Coherence Loss: All Objects Matter in Salient and Camouflaged Object Detection

arXiv

引用

arXiv 2024年

作者： Yang, Ziyun Choy, Kevin Farsiu, Sina The Biomedical Engineering Department Duke University DurhamNC27705 United States The Biomedical Engineering Electrical and Computer Engineering Ophthalmology and Computer Science Departments Duke University DurhamNC27705 United States

Generic object detection is a category-independent task that relies on accurate modeling of objectness. We show that for accurate semantic analysis, the network needs to learn all object-level predictions that appear at any stage of learning, including the pre-defined ground truth (GT) objects and the ambiguous decoy objects that the network misidentifies as foreground. Yet, most relevant models focused mainly on improving the learning of the GT objects. A few methods that consider decoy objects utilize loss functions that only focus on the single-response, i.e., the loss response of a single ambiguous pixel, and thus do not benefit from the wealth of information that an object-level ambiguity learning design can provide. Inspired by the human visual system, which first discerns the boundaries of ambiguous regions before delving into the semantic meaning, we propose a novel loss function, Spatial Coherence Loss (SCLoss), that incorporates the mutual response between adjacent pixels into the widely-used single-response loss functions. We demonstrate that the proposed SCLoss can gradually learn the ambiguous regions by detecting and emphasizing their boundaries in a self-adaptive manner. Through comprehensive experiments, we demonstrate that replacing popular loss functions with SCLoss can improve the performance of current state-of-the-art (SOTA) salient or camouflaged object detection (SOD or COD) models. We also demonstrate that combining SCLoss with other loss functions can further improve performance and result in SOTA outcomes for different applications. The codes will be released to https://***/TBD upon acceptance. Copyright © 2024, The Authors. All rights reserved.

关键词： Object recognition

来源：评论

学校读者我要写书评

暂无评论

Balancing memorization and generalization in RNNs for high performance brain-machine interfaces 23

Balancing memorization and generalization in RNNs for high p...

引用

Proceedings of the 37th International Conference on Neural Information Processing Systems

作者： Joseph T. Costello Hisham Temmar Luis H. Cubillos Matthew J. Mender Dylan M. Wallace Matthew S. Willsey Parag G. Patil Cynthia A. Chestek Departments of Electrical and Computer Engineering Biomedical Engineering Robotics and Neurosurgery University of Michigan Ann Arbor MI

Brain-machine interfaces (BMIs) can restore motor function to people with paralysis but are currently limited by the accuracy of real-time decoding algorithms. Recurrent neural networks (RNNs) using modern training techniques have shown promise in accurately predicting movements from neural signals but have yet to be rigorously evaluated against other decoding algorithms in a closed-loop setting. Here we compared RNNs to other neural network architectures in real-time, continuous decoding of finger movements using intracortical signals from nonhuman primates. Across one and two finger online tasks, LSTMs (a type of RNN) outperformed convolutional and transformer-based neural networks, averaging 18% higher throughput than the convolution network. On simplified tasks with a reduced movement set, RNN decoders were allowed to memorize movement patterns and matched able-bodied control. Performance gradually dropped as the number of distinct movements increased but did not go below fully continuous decoder performance. Finally, in a two-finger task where one degree-of-freedom had poor input signals, we recovered functional control using RNNs trained to act both like a movement classifier and continuous decoder. Our results suggest that RNNs can enable functional real-time BMI control by learning and generating accurate movement patterns.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Recent advances in microsystem approaches for mechanical characterization of soft biological tissues

引用

Microsystems & Nanoengineering 2022年第4期8卷 1-16页

作者： Enming Song Ya Huang Ningge Huang Yongfeng Mei Xinge Yu John A.Rogers Shanghai Frontiers Science Research Base of Intelligent Optoelectronics and Perception Institute of OptoelectronicsFudan UniversityShanghai 200433China International Institute of Intelligent Nanorobots and Nanosystems Fudan UniversityShanghai 200433China Department of Biomedical Engineering City University of Hong KongHong Kong 999077China Department of Materials Science Fudan UniversityShanghai 200433China Querrey Simpson Institute for Bioelectronics Department of Materials Science and EngineeringDepartments of Biomedical EngineeringNeurological SurgeryChemistryMechanical EngineeringElectrical Engineering and Computer ScienceNorthwestern UniversityEvanstonIL 60208USA

Microsystem technologies for evaluating the mechanical properties of soft biological tissues offer various capabilities relevant to medical research and clinical diagnosis of pathophysiologic *** progress includes(1)the development of tissue-compliant designs that provide minimally invasive interfaces to soft,dynamic biological surfaces and(2)improvements in options for assessments of elastic moduli at spatial scales from cellular resolution to macroscopic areas and across depths from superficial levels to deep *** review summarizes a collection of these technologies,with an emphasis on operational principles,fabrication methods,device designs,integration schemes,and measurement *** core content begins with a discussion of platforms ranging from penetrating filamentary probes and shape-conformal sheets to stretchable arrays of ultrasonic *** sections examine different techniques based on planar microelectromechanical system(MEMS)approaches for biocompatible interfaces to targets that span scales from individual cells to *** highlighted example includes miniature electromechanical devices that allow depth profiling of soft tissue biomechanics across a wide range of *** clinical utility of these technologies is in monitoring changes in tissue properties and in targeting/identifying diseased tissues with distinct variations in *** results suggest future opportunities in engineered systems for biomechanical sensing,spanning a broad scope of applications with relevance to many aspects of health care and biology research.

关键词： advances identifying utility

来源：评论

学校读者我要写书评

暂无评论

Sparsity-Aware Hardware-Software Co-Design of Spiking Neural Networks: An Overview

Sparsity-Aware Hardware-Software Co-Design of Spiking Neural...

引用

IEEE International Symposium on Embedded Multicore Socs (MCSoC)

作者： Ilkin Aliyev Kama Svoboda Tosiron Adegbija Jean-Marc Fellous Department of Electrical & Computer Engineering University of Arizona Tucson AZ USA Departments of Psychology and Biomedical Engineering University of Arizona Tucson AZ USA

ISBN: (数字)9798331530471

ISBN: (纸本)9798331530488

关键词： Training Heuristic algorithms Computational modeling Neurons Spiking neural networks computer architecture Hardware Energy efficiency Encoding Artificial intelligence

来源：评论

学校读者我要写书评

暂无评论

Joint Source Decomposition of Concurrent EEG-FMRI Data in Epilepsy and Control Groups 22

Joint Source Decomposition of Concurrent EEG-FMRI Data in Ep...

引用

22nd IEEE International Symposium on biomedical Imaging, ISBI 2025

作者： Harding, J. Mason Pourmotabbed, Haatef Li, Yamin Rogge-Obando, Kimberly Wang, Kate Goodale, Sarah E. Wang, Shiyu Bibro, Camden Allee, Bergen Martin, Caroline Morgan, Victoria L. Englot, Dario J. Chang, Catie Departments of Electrical and Computer Engineering Vanderbilt University United States Departments of Biomedical Engineering Vanderbilt University United States Departments of Computer Science Vanderbilt University United States Neuroscience Graduate Program Vanderbilt University United States Departments of Radiology Vanderbilt University Medical Center United States Departments of Neurological Surgery Vanderbilt University Medical Center United States Vanderbilt Memory and Alzheimer's Center Vanderbilt University Medical Center United States

ISBN: (纸本)9798331520526

EEG and fMRI are complementary, noninvasive technologies for investigating human brain function. These modalities have been used to uncover large-scale functional networks and their disruptions in clinical populations. Given the high temporal resolution of EEG and high spatial resolution of fMRI, integrating these modalities can provide a more holistic understanding of brain activity. This work explores a multimodal source decomposition technique for extracting shared modes of temporal variation between fMRI BOLD signals and EEG spectral power fluctuations in the resting state. The resulting components are then compared between patients with focal epilepsy and controls, revealing multimodal network differences between groups. © 2025 IEEE.

关键词： EEG fMRI functional connectivity ICA neuroimaging

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：