检索结果-内蒙古大学图书馆

3rd BenchCouncil International Symposium on Benchmarking, Measuring, and Optimizing, Bench 2020

作者： Chen, Bangduo Li, Mingzhen Yang, Hailong Luan, Zhongzhi Gan, Lin Yang, Guangwen Qian, Depei School of Computer Science and Engineering Beihang University Beijing100191 China State Key Laboratory of Mathematical Engineering and Advanced Computing Wuxi Jiangsu214125 China Department of Computer Science and Technology Tsinghua University Beijing100084 China

ISBN: (纸本)9783030710576

The Sunway processor has been demonstrated with superior performance by various scientific applications, domain specific frameworks and numerical algorithms. However, the optimization techniques that can fully exploit the architecture features are usually buried deep in large code bases, which prevents average programmers to understand such optimization techniques. Thus, the existing complex software fails to provide guidance for more programs embracing the computation power of Sunway processor. In this paper, we build a benchmark suite swRodinia by porting and optimizing the well-known Rodinia benchmark on Sunway processor. Specifically, we demonstrate several optimization techniques by tailoring the benchmarks to better leverage the architecture features for higher performance. Moreover, based on the optimization experiences, we derive several useful insights from both software and hardware perspectives, that not only guide the better utilization of current Sunway processor, but also reveal the direction of hardware improvements for future Sunway processor. We open source the swRodinia benchmark suite and encourage the community to enhance the benchmark with us continuously. © 2021, Springer Nature Switzerland AG.

关键词： Open source software

来源：评论

学校读者我要写书评

暂无评论

HELCFL: high-efficiency and low-cost federated learning in heterogeneous mobile-edge computing 22

HELCFL: high-efficiency and low-cost federated learning in h...

引用

Proceedings of the 2022 Conference & Exhibition on Design, Automation & Test in Europe

作者： Yangguang Cui Kun Cao Junlong Zhou Tongquan Wei East China Normal University Shanghai China and State Key Laboratory of Mathematical Engineering and Advanced Computing Wuxi China Jinan University Guangzhou China Nanjing University of Science and Technology Nanjing China

ISBN: (纸本)9783981926361

Federated Learning (FL), an emerging distributed machine learning (ML), empowers a large number of embedded devices (e.g., phones and cameras) and a server to jointly train a global ML model without centralizing user private data on a server. However, when deploying FL in a mobile-edge computing (MEC) system, restricted communication resources of the MEC system, heterogeneity and constrained energy of user devices have a severe impact on FL training efficiency. To address these issues, in this article, we design a distinctive FL framework, called HELCFL, to achieve high-efficiency and low-cost FL training. Specifically, by analyzing the theoretical foundation of FL, our HELCFL first develops a utility-driven and greedy-decay user selection strategy to enhance FL performance and reduce training delay. Subsequently, by analyzing and utilizing the slack time in FL training, our HELCFL introduces a device operating frequency determination approach to reduce training energy costs. Experiments verify that our HELCFL can enhance the highest accuracy by up to 43.45%, realize the training speedup of up to 275.03%, and save up to 58.25% training energy costs compared to state-of-the-art baselines.

关键词： high efficiency user selection federated learning low costs mobile-edge computing frequency determination

来源：评论

学校读者我要写书评

暂无评论

Progressive Multi-Stage Feature Mix for Person Re-Identification

Progressive Multi-Stage Feature Mix for Person Re-Identifica...

引用

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

作者： Yan Zhang Binyu He Li Sun Qingli Li Shanghai Key Laboratory of Multidimensional Information Processing East China Normal University Shanghai China State Key Laboratory of Mathematical Engineering and Advanced Computing Wuxi China

Image features from a small local region often give strong evidence in person re-identification task. However, CNN suffers from paying too much attention on the most salient local areas, thus ignoring other discriminative clues, e.g., hair, shoes or logos on clothes. In this work, we propose a Progressive Multi-stage feature Mix network (PMM), which enables the model to find out the more precise and diverse features in a progressive manner. Specifically, (i) to enforce the model to look for different clues in the image, we adopt a multi-stage classifier and expect that the model is able to focus on a complementary region in each stage. (ii) we propose an Attentive feature Hard-Mix (A-Hard-Mix) to replace the salient feature blocks by the negative example in the current batch, whose label is different from the current sample. (iii) extensive experiments have been carried out on reID datasets such as the Market-1501, DukeMTMC-reID and CUHK03, showing that the proposed method can boost the re-identification performance significantly. Source code¹ has been released.

关键词： Training Hair Conferences Footwear Signal processing Acoustics Task analysis

来源：评论

学校读者我要写书评

暂无评论

Alleviating Imbalance in Synchronous Distributed Training of Deep Neural Networks

Alleviating Imbalance in Synchronous Distributed Training of...

引用

IEEE International Conference on Big Data and Cloud computing (BdCloud)

作者： Haiyang Lin Mingyu Yan Duo Wang Wenming Li Xiaochun Ye Zhimin Tang Dongrui Fan State Key Laboratory of Computer Architecture ICT CAS Beijing China University of Chinese Academy of Sciences Beijing China State Key Laboratory of Mathematical Engineering and Advanced Computing Wuxi China

Deep neural networks achieve huge success in computer vision, natural language process and other fields. As larger networks and larger datasets bring better performance, the training time grows rapidly. Distributed training uses parallelization strategy to accelerate the training of deep neural network. In the existing distributed training framework of deep neural network, the parameter server architecture has poor scalability and is not suitable for large-scale distributed training. Ring/HD All Reduce algorithm perform well in scalability under ideal circumstances. However, in many realistic scenarios, imbalance will occur frequently in the distributed training system, which is mainly the imbalance of compute ability and network. Due to the structure characteristics of synchronous distributed training and All Reduce algorithm, imbalance will bring serious impact on performance. At present, there is a lack of research on both compute ability imbalance and network imbalance, especially dynamic imbalance. To alleviate this problem, we analyze All Reduce algorithm first, and propose RingDL, which optimizes the performance of distributed training system by alleviating imbalance through customizable dynamic workloads strategy and link strategy. The experimental results show that RingDL significantly reduces idle time in computing phase and synchronous phase, improving the overall performance.

关键词： Training Deep learning Degradation Computer vision Heuristic algorithms Scalability Neural networks

来源：评论

学校读者我要写书评

暂无评论

The various graphs in graph computing 1

引用

22nd Annual Conference on Computer engineering and Technology, NCCET 2018

作者： Sun, Rujun Zhang, Lufei State Key Laboratory of Mathematical Engineering and Advanced Computing Wuxi China

ISBN: (数字)9789811359194

ISBN: (纸本)9789811359187

The world is full of relationships, and graph is the most evident representation for them. With the increasing of data scale, graphs become larger and have encountered a new world of analyzing. What can we learn from a graph? How many kinds of graphs are there? How different is graph from one area to that from another? All these questions need answers, but previous research on graph computing mainly focused on computing frameworks and systems, paying little attention to graph itself. In this paper, we studied graphs of different kinds, different scales and different mining methods, trying to give a sketcher and classification of graph categories. Besides, we studied characters and analyzed algorithms in each category. We researched public graph datasets to show current graph scale and its trend for future infrastructure. © 2019, Springer Nature Singapore Pte Ltd.

关键词： Graphic methods

来源：评论

学校读者我要写书评

暂无评论

Active Defense Technology Based on Natural Software Diversity in Java Web Services 4

Active Defense Technology Based on Natural Software Diversit...

引用

2020 4th International Workshop on advanced Algorithms and Control engineering, IWAACE 2020

作者： Liu, Zhenwu Zhang, Zheng Liu, Hao Zhang, Gaofei Chen, Yuan State Key Laboratory of Mathematical Engineering and Advanced Computing ZhengZhou HeNan450001 China PLA Strategic Support Force Information Engineering University ZhengZhou HeNan450001 China Jiangnan Institute of Computing Technology WuXi JiangSu214083 China

With the popular use of network application services and the growing number of users, web services have become the main target of hackers. Traditional single software has insufficient security protection against unknown threats and is prone to various vulnerabilities. In this paper, we adopt a new type of active defense strategy in java web services. At different application layers, we use existing components based on the natural software diversity to form different software stacks, and use dynamic scheduling mechanisms to change the attack surface at all times to obtain security protection. Our analysis shows that the flexibility of java web services using this defense strategy is enhanced, and the exploitability of its vulnerabilities is effectively reduced. © Published under licence by IOP Publishing Ltd.

关键词： Web services

来源：评论

学校读者我要写书评

暂无评论

Research on 3D Human Body Reconstruction Algorithm of Monocular Image

Research on 3D Human Body Reconstruction Algorithm of Monocu...

引用

Artificial Intelligence Technology (ACAIT), Asian Conference on

作者： Zijian Liang Huan Liang Wenlong Fu Neuroscience and Intelligent Media Institute Communication University of China Beijing China State Key Laboratory of Media Convergence and Communication Communication University of China Beijing China State Key Laboratory of Mathematical Engineering and Advanced Computing Wuxi China

ISBN: (纸本)9781665426312

The three-dimensional human body feature parameters extracted by the neural network method are coupled in the high-dimensional feature space. The fully connected network processes the three-dimensional human pose and shape feature information synchronously, which cannot handle multifactor features well, which makes the model perform poorly. To solve this problem, on the basis of the benchmark network, a decoupling model is added, which processes the three-dimensional human pose and shape information separately, and finally integrates the features through the bi-linear transformation method. Through experimental verification, the algorithm effectively reduces the reconstruction error, and the improved algorithm performs better.

关键词： Couplings Solid modeling Three-dimensional displays Shape Biological system modeling Neural networks Reconstruction algorithms

来源：评论

学校读者我要写书评

暂无评论

Implementation of Multi-compartment Neuron Model with Synaptic Plasticity Using FPGA

Implementation of Multi-compartment Neuron Model with Synapt...

引用

International Conference on Communication Technology (ICCT)

作者： Yiwen Ding Wenlong Fu Lihong Cao Neuroscience and Intelligent Media Institute Communication University of China Beijing China State Key Laboratory of Media Convergence and Communication Communication University of China Beijing China State Key Laboratory of Mathematical Engineering and Advanced Computing Wuxi China

ISBN: (纸本)9781665432078

Computational neuroscience is interdisciplinary and plays an essential role in facilitating the development of cognitive neuroscience and Artificial General Intelligence (AGI). To explore the computational efficiency of Spiking Neural Network (SNN) on a computational hardware platform, we deploy SNN on the Xilinx Zynq7 Zedboard Field Programmable Gate Array (FPGA). Specifically, we use multi-compartmental Hodgkin-Huxley (HH) neuron model to construct neural circuits via C++ code and adopt a biologically plausible calcium concentration-based plasticity model as the learning rule of SNN. Our experimental results show that the FPGA platform can support parallel computation of the multi-compartment HH model with synaptic plasticity. Compared to other methods, FPGA-based computing has a significant speed advantage, suggesting that FPGA can be a good choice for implementing brain-inspired computation.

关键词： Computational modeling Biological system modeling Neurons Neural circuits Logic gates Brain modeling Hardware

来源：评论

学校读者我要写书评

暂无评论

A Survey on Neural Machine Translation Applied to Sign Language Generation

A Survey on Neural Machine Translation Applied to Sign Langu...

引用

International Conference on Applied Machine Learning (ICAML)

作者： Yue Zhang Lihong Cao State Key Laboratory of Media Convergence and Communication Communication University of China Beijing China Neuroscience and Intelligent Media Institute Communication University of China Beijing China State Key Laboratory of Mathematical Engineering and Advanced Computing Wuxi China

ISBN: (纸本)9781665421263

Sign languages are the primary medium of communication for the deaf and hearing-impaired community, and the number of human beings with hearing impairment is increasing at a rapid rate. There is a significant communication barrier between the hearing-impaired and normal-hearing communities. One solution to this issue is to offload the effort of text understanding for the deaf and hearing-impaired community in the form of Sign Language Generation (SLG). In this paper, first of all, we systematically review the mainstream Neural Machine Translation (NMT)-based SLG approaches. Specifically, according to their model architectures, we divide them into two categories: the Recurrent Neural Networks (RNNs)-based and Transformer-based models, and discuss their advantages and limitations. Next, we introduce the existing publicly available datasets and the data preprossessing methods in the field of SLG. Then, we summarize the performance evaluation methods for NMT-based SLG models. Finally, we offer two possible future research directions for better development in the field of SLG.

关键词： Performance evaluation Recurrent neural networks Pipelines Data preprocessing Gesture recognition Machine learning Auditory system

来源：评论

学校读者我要写书评

暂无评论

Empirical investigation of stochastic local search for maximum satisfiability

引用

Frontiers of Computer Science 2019年第1期13卷 86-98页

作者： Yi CHU Chuan LUO Shaowei CAI Haihang YOU State Key Laboratory of Computer Architecture Institute of Computing TechnologyChinese Academy of Sciences Beijing 100190China University of Chinese Academy of Sciences Beijing 100049China State Key Laboratory of Mathematical Engineering and Advanced Computing Wuxi 214125China State Key Laboratory of Computer Science Institute of SoftwareChinese Academy of SciencesBeijing 100190China

The maximum satisfiability (MAX-SAT)problem is an important NP-hard problem in theory,and has a broad range of applications in *** local search (SLS)is becoming an increasingly popular method for solving ***,a powerful SLS algorithm called CCLS shows efficiency on solving random and crafted MAX-SAT ***,the performance of CCLS on solving industrial MAX-SAT instances lags far *** this paper,we focus on experimentally analyzing the performance of SLS algorithms for solving industrial MAXSAT ***,we conduct experiments to analyze why CCLS performs poor on industrial *** we propose a new strategy called additive BMS (Best from Multiple Selections)to ease the serious *** integrating CCLS and additive BMS,we develop a new SLS algorithm for MAXSAT called CCABMS,and related experiments indicate the efficiency of ***,we experimentally analyze the effectiveness of initialization methods on SLS algorithms for MAX-SAT,and combine an effective initialization method with CCABMS,resulting in an enhanced *** results show that our enhanced algorithm performs better than its state-of-the-art SLS competitors on a large number of industrial MAX-SAT instances.

关键词： empirical investigation stochastic local search maximum satisfiability industrial instances additive BMS

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：