检索结果-内蒙古大学图书馆

arXiv 2024年

作者： Li, Hongkang Wang, Meng Zhang, Shuai Liu, Sijia Chen, Pin-Yu Dept. Electrical Computer and System Engineering Rensselaer Polytechnic Institute TroyNY United States Dept. Data Science New Jersey Institute of Technology NewarkNJ United States Dept. Computer Science and Engineering Michigan State University East LansingMI United States IBM Thomas J. Watson Research Center Yorktown HeightsNY United States

Efficient training and inference algorithms, such as low-rank adaption and model pruning, have shown impressive performance for learning Transformer-based large foundation models. However, due to the technical challenges of the non-convex optimization caused by the complicated architecture of Transformers, the theoretical study of why these methods can be applied to learn Transformers is mostly elusive. To the best of our knowledge, this paper shows the first theoretical analysis of the property of low-rank and sparsity of one-layer Transformers by characterizing the trained model after convergence using stochastic gradient descent. By focusing on a data model based on label-relevant and label-irrelevant patterns, we quantify that the gradient updates of trainable parameters are low-rank, which depends on the number of label-relevant patterns. We also analyze how model pruning affects the generalization while improving computation efficiency and conclude that proper magnitude-based pruning has a slight effect on the testing performance. We implement numerical experiments to support our findings. Copyright © 2024, The Authors. All rights reserved.

关键词： Stochastic systems

来源：评论

学校读者我要写书评

暂无评论

Self-Supervised Motion Segmentation with Confidence-Aware Loss Functions for Handling Occluded Pixels and Uncertain Optical Flow Predictions

Self-Supervised Motion Segmentation with Confidence-Aware Lo...

引用

IEEE/RSJ International Conference on Intelligent Robots and systems (IROS)

作者： Chung-Yu Chen Bo-Yun Lai Ying-Shiuan Huang Wen-Chieh Lin Chieh-Chih Wang Dept. of Computer Science National Yang Ming Chiao Tung Univ. College of Electrical and Computer Engineering National Yang Ming Chiao Tung University Hsinchu Taiwan Mechanical and Mechatronics Systems Research Laboratories Industrial Technology Research Institute Hsinchu Taiwan

ISBN: (数字)9798350377705

ISBN: (纸本)9798350377712

In driving scenarios, motion segmentation is a crucial and fundamental component that is needed for many tasks. Recently, a self-supervised multitasking framework was proposed for driving scenarios. It simultaneously trains motion segmentation, optical flow, dept., and ego-motion models without annotated data. The self-supervised architecture derives training signals from training data via loss functions. If these loss functions lack robustness, they may result in model inaccuracies. To reduce the bad influences of occlusion and optical flow estimation errors on motion segmentation, we propose two loss functions: (1) Soft-Per-Pixel-Minimum (Soft-PPM) loss that excludes occluded pixels while balancing the contribution of each frame on the loss function temporally; (2) Flow difference loss that excludes pixels with unclear motion states to diminish the effect of optical flow estimation errors. Our loss function design is based on the key insight that information such as dept. and optical flow can be used to train motion segmentation models and act as a reliable measure for pixels during training. Our approach can improve segmentation accuracy for both moving and static objects and has achieved IoU scores on moving and static classes comparable to the state-of-the-art methods on the KITTI dataset.

关键词： Optical losses Training computer vision Estimation error Motion segmentation Training data Loss measurement Data models Robustness Optical flow

来源：评论

学校读者我要写书评

暂无评论

Near field radio link design in fat phantom for Wireless Capsule Endoscopy application

Near field radio link design in fat phantom for Wireless Cap...

引用

2022 Malaysia-Japan Workshop on Radio technology, MJWRT 2022

作者： Kamaruddin, Nur Amalina Kamardin, Kamilia Yamada, Yoshihide Zainudin, Norsiha Arima, Takuji Michishita, Naobumi Takahashi, Masaharu Communication Systems and Network IKohza Department of Electronic Systems Engineering Malaysia-Japan International Institute of Technology Universiti Teknologi Malaysia Kuala Lumpur54100 Malaysia Department of Electrical Engineering Faculty of Engineering University of Malaya Kuala Lumpur50603 Malaysia Dept. of Electronic and Information Engineering Tokyo University of Agriculture and Technology Koganei Tokyo184-8588 Japan National Defense Academy Yokosuka Japan Center for Frontier Medical Engineering Chiba University Chiba Japan

Medical devices such as Wireless Capsule Endoscopy (WCE) that need wireless imaging and monitoring, operating at 402 MHz, is considered to be in the near field region due to the distance being shorter than the wavelength. In this paper, the radio link calculation method in the near field region is investigated. Normal Mode Helical Antenna (NMHA) is used as the transmitter antenna which is placed inside of a fat phantom of 11.6 F/m and the receiver antenna at distance 0.1 to 1.0 m apart. For link budget establishment, parameters such as power density, W and receiver antenna effective area, Ae are investigated through electromagnetic simulation. The distance dependency of W is examined and the result of W 1/(distance)2 is obtained. The relation of Ae of the receiver antenna is found to be the size of the cross-section area of the antenna. A radio link equation is formed by employing W, Ae and other propagation loss factors of a medium and is then compared with the power received, PR and S21 obtained from simulation. The agreement between calculated and simulated results shows that this equation can be used in the near field region. © Published under licence by IOP Publishing Ltd.

关键词： Receiving antennas

来源：评论

学校读者我要写书评

暂无评论

Using the recurrence plots as indicators for the recognition of Parkinson’s disease through phonemes assessment

Using the recurrence plots as indicators for the recognition...

引用

Annual International Conference of the IEEE engineering in Medicine and Biology Society (EMBC)

作者： Vasileios Skaramagkas Anastasia Pentari Dimitrios I. Fotiadis Manolis Tsiknakis Dept. of Electrical and Computer Engineering Hellenic Mediterranean University Heraklion Crete Greece Institute of Computer Science Foundation for Research and Technology Hellas (FORTH) Heraklion Crete Greece Dept. of Biomedical Research Institute of Molecular Biology and Biotechnology (FORTH) Ioannina Greece Dept. of Materials Science and Engineering Unit of Medical Technology and Intelligent Information Systems University of Ioannina Ioannina Greece

Parkinson’s disease (PD) is considered to be the second most common neurodegenerative disease which affects the patients’ life throughout the years. As a consequence, its early diagnosis is of major importance for the improvement of life quality, implying that the severe symptoms can be delayed through appropriate clinical intervention and treatment. Among the most important premature symptoms of PD are the voice impairments of articulation, phonation and prosody. The objective of this study is to investigate whether the voice’s dynamic behavior can be used as possible indicator for PD. Thus in this work, we employ the recurrence plots (RPs) which derive from the analysis of the three modulated vowels /a/, /e/ and /o/, which belong to the PC-GITA dataset, and are fed as input images to a 3-channel Convolutional Neural Network-based (CNN) architecture, which, finally, differentiates the 50 PD patients from 50 healthy subjects. The experimental results obtained provide evidence that the RP-based approach is a promising tool for the recognition of PD patients through the analysis of voice recordings, with a classification accuracy achieved equal to 87%.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Modified Convolutional Neural Network Algorithm for Categorization of Solar PV Panel Soiling

Modified Convolutional Neural Network Algorithm for Categori...

引用

Smart Technologies for Power and Renewable Energy (SPECon), International Conference on

作者： Abhyudoy Chaki Rachna Bhatia Pallavi Mishra R. Raja Singh N. Rajasekar School of Computer Science Vellore Institute of Technology Vellore India Dept. of Mathematics School of Advanced Sciences Vellore Institute of Technology Vellore India Automotive Research Center Vellore Institute of Technology Vellore India School of Electrical Engineering Vellore Institute of Technology Vellore India

ISBN: (数字)9798350383041

ISBN: (纸本)9798350383058

The identification and categorization of soiling using artificial intelligence is essential for improving solar photovoltaic (PV) system efficiency. It is possible to use focused and effective cleaning techniques on solar panels by classifying and comprehending the particular types of pollutants. This research proposes a modified convolutional neural network (CNN) approach to efficiently classify soiling on solar PV panels. In contrast to the conventional method, the suggested CNN model is created using simple images rather than images plus time series data. As a result, processing speed is increased. Additionally, the suggested model achieved 96.91 percent accuracy. This success establishes our suggested methodology as a ground-breaking development in soiling type detection, providing improved accuracy and a unique strategy in contrast to current approaches. The trained and tested model is deployed on the web using the Flask. Flask is a powerful and flexible microweb framework for Python, Finally, a comparison is made between the existing method and the proposed method.

关键词： Renewable energy sources Time series analysis Data models Cleaning Classification algorithms Convolutional neural networks Solar panels

来源：评论

学校读者我要写书评

暂无评论

Sparsity-Smoothness-Aware Power Spectral Density Estimation with Application to Phased Array Weather Radar

Sparsity-Smoothness-Aware Power Spectral Density Estimation ...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Hiroki Kuroda Daichi Kitahara Eiichi Yoshikawa Hiroshi Kikuchi Tomoo Ushio Dept. of Information and Management Systems Engineering Nagaoka University of Technology Division of Electrical Electronic and Infocommunications Engineering Osaka University Aeronautical Technology Directorate Japan Aerospace Exploration Agency Center for Space Science and Radio Engineering The University of Electro-Communications

In this paper, we propose a sparsity and smoothness regularized model for the estimation of nonnegative power spectral densities (PSDs) of complex-valued random processes from mixtures of realizations. The proposed model is designed to jointly estimate frequency components of the realizations and the PSDs. The PSDs are estimated by the nonnegative variable in the proposed model, which enables that the sparsity and the smoothness can be exploited via convex optimization. Numerical experiments on the phased array weather radar, which is an advanced weather radar system, show that the proposed approach achieves superior performance to the existing sparse estimation models combined with post-smoothing.

关键词： Phased arrays Meteorological radar Estimation Frequency estimation Radar signal processing Numerical models Random processes

来源：评论

学校读者我要写书评

暂无评论

Robust Charging Solutions for 400V EV Batteries: Implementing a Quasi Single-Stage Current-Fed Resonant Converter

Robust Charging Solutions for 400V EV Batteries: Implementin...

引用

IEEE Texas Power and Energy Conference (TPEC)

作者： Muhammad Furqan Waqar Uddin Kamran Zeb Muhammad Ashar Ayaz Irfan Khan Muhammad Khalid School of Electrical Engineering and Computer Science (SEECS) National University of Science and Technology (NUST) Islamabad Pakistan Interdisciplinary Research Center for Sustainable Energy Systems (IRC-SES) King Fahd University of Petroleum and Minerals (KFUPM) Dhahran Saudi Arabia IRC-SES KFUPM Dhahran Saudi Arabia Dept. of Electrical and Computer Engneering Texas A and M University College Station TX USA Dept. of Elec. and Comp. Engineering Texas A and M University College Station TX USA Electrical Engineering Department KFUPM and IRC-SES KFUPM Dhahran Saudi Arabia

ISBN: (数字)9798331541125

ISBN: (纸本)9798331541132

This work presents a high-performance AC-DC quasi single-stage current-fed resonant converter based electric vehicles (EVs) charger. A non-linear control technique is applied which has two main objectives i.e; Power factor correction (PFC) and regulation of DC output voltage. For this purpose, super twisting sliding mode control (ST-SMC) is suggested in this study for 400V batteries of EVs. To reduce switching losses and improve efficiency, the secondary side switch shows turn OFF with zero voltage switching (ZVS) while the primary ones exhibit ZVS at turn ON instant. Hardware-in-loop (HIL) experimental validation based on the Delfino F28369D dual-core microcontroller is used to verify the performance of the controller and converter simulated on PSIM and MATLAB/Simulink.

关键词： Microcontrollers Switching loss Switches Resonant converters Zero voltage switching Regulation Batteries Safety Voltage control Sliding mode control

来源：评论

学校读者我要写书评

暂无评论

Multi-Person Key Points Detection for Abnormal Human Behavior Analysis Using The ConvLSTM-AE Method

Multi-Person Key Points Detection for Abnormal Human Behavio...

引用

International Conference on computer engineering, Network and Intelligent Multimedia (CENIM)

作者： Sofia Ariyani Eko Mulyanto Yuniarno Mauridhi Hery Purnomo Dept. Of Electrical Engineering Dept. of Electrical Engineering Faculty of Intelligent Electrical and Informatics Technology (F-Electics) Faculty of engineering Institut Teknologi Sepuluh Nopember Universitas Muhammadiyah Jember Surabaya Indonesia Dept. Of Electrical Engineering Dept. of Computer Engineering Faculty of Intelligent Electrical and Informatics Technology (F-Electics) Institut Teknologi Sepuluh Nopember Surabaya Indonesia Department of Electrical Engineering Dept. of Computer Engineering Faculty of Intelligent Electrical and Informatics Technology (F-Electics) Institut Teknologi Sepuluh Nopember University Center of Excellence on Artificial Intelligence for Healthcare and Society Surabaya Indonesia

ISBN: (纸本)9781665476515

The detection of abnormal human behavior is an interesting issue to consider in computer vision. The problem of detecting abnormal activities carried out in this study is a problem formulated in the process of monitoring human activities. In understanding the nature of human activities, a system is needed that is applied to data training which is specifically proposed based on ConvLSTM-AE for the detection of moving objects in motion-based events. Detection of anomalies and localization that arise due to camera jitter and movement of objects in the area. With the unusual motion of the video frame, connected component analysis is exploited to provide bounding boxing to find out whether human movements whose activities are based on human poses will be classified as normal or abnormal in a set of activity data. The learning and training process so that the proposed model can capture different duration of time and more optimal processes.

关键词： Training Location awareness computer vision Surveillance Multimedia systems Computational modeling Estimation

来源：评论

学校读者我要写书评

暂无评论

ACT-Net: Anchor-context Action Detection in Surgery Videos

arXiv

引用

arXiv 2023年

作者： Hao, Luoying Hu, Yan Lin, Wenjun Wang, Qun Li, Heng Fu, Huazhu Duan, Jinming Liu, Jiang School of Computer Science University of Birmingham United Kingdom Research Institute of Trustworthy Autonomous Systems Dept. of Computer Science and Engineering Southern University of Science and Technology China Dept. of Mechanical Engineering National University of Singapore Singapore Third medical center of Chinese PLAGH China Singapore

Recognition and localization of surgical detailed actions is an essential component of developing a context-aware decision support system. However, most existing detection algorithms fail to provide high-accuracy action classes even having their locations, as they do not consider the surgery procedure’s regularity in the whole video. This limitation hinders their application. Moreover, implementing the predictions in clinical applications seriously needs to convey model confidence to earn entrustment, which is unexplored in surgical action prediction. In this paper, to accurately detect fine-grained actions that happen at every moment, we propose an anchor-context action detection network (ACTNet), including an anchor-context detection (ACD) module and a class conditional diffusion (CCD) module, to answer the following questions: 1) where the actions happen;2) what actions are;3) how confidence predictions are. Specifically, the proposed ACD module spatially and temporally highlights the regions interacting with the extracted anchor in surgery video, which outputs action location and its class distribution based on anchor-context interactions. Considering the full distribution of action classes in videos, the CCD module adopts a denoising diffusion-based generative model conditioned on our ACD estimator to further reconstruct accurately the action predictions. Moreover, we utilize the stochastic nature of the diffusion model outputs to access model confidence for each prediction. Our method reports the state-of-the-art performance, with improvements of 4.0% mAP against baseline on the surgical video dataset. Copyright © 2023, The Authors. All rights reserved.

关键词： Diffusion

来源：评论

学校读者我要写书评

暂无评论

Optimization of Imprecise Multiplier Circuits by using Binary Decision Diagram

Optimization of Imprecise Multiplier Circuits by using Binar...

引用

IEEE International Symposium on Smart Electronic systems (iSES)

作者： Pooja Choudhary Lava Bhargava G.U. Vinod Ashok Kumar Suhag Mashahiro Fujita Virendra Singh Dept. of Electronics and Communication Engineering Swami Keshvanand Institute of Technology Management and Gramothan Jaipur India Malaviya National Institute of Technology Jaipur India Dept. of Electronics and Communication Engineering Malaviya National Institute of Technology Jaipur India Qualcomm Bangalore India Dept. of Electronics and Communication Engineering BML Munjal University Haryana India VLSI Design and Education Center University of Tokyo Japan Dept. of Electrical Engineering & Dept. of Computer Science Engineering Indian Institute of Technology Bombay India

Imprecise multipliers are employed to enhance the efficiency and reduce the power consumption of arithmetic circuits while tolerating acceptable levels of inaccuracy in the outcomes. The multiplier circuit has a broader range of applications within embedded systems, particularly in the fields of digital signal processing, IoT and multimedia, etc. We proposed a systematic formal approach utilizing the Look-Up Table (LUT) employing the Burch method to evaluate the imprecise multiplier based on the Binary Decision Diagram (BDD) considering Worst Case Error constraint. Implemented 13 bit multiplier using normal and proposed imprecise multiplier shows 98% reduction in number of nodes. The proposed 25 bit imprecise multiplier based on Burch HL shows an 89% decrement in a number of nodes w.r.t. Burch LH method. BDD generation for imprecise multiplier is 75% faster than normal.

关键词：

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：