检索结果-内蒙古大学图书馆

arXiv 2020年

作者： Gao, Wanling Tang, Fei Zhan, Jianfeng Lan, Chuanxin Luo, Chunjie Wang, Lei Dai, Jiahui Cao, Zheng Xiong, Xiongwang Jiang, Zihan Hao, Tianshu Fan, Fanda Wen, Xu Zhang, Fan Huang, Yunyou Chen, Jianan Du, Mengjia Ren, Rui Zheng, Chen Zheng, Daoyi Tang, Haoning Zhan, Kunlin Wang, Biao Kong, Defei Yu, Minghe Tan, Chongkang Li, Huan Tian, Xinhui Li, Yatao Lu, Gang Shao, Junchao Wang, Zhenyu Wang, Xiaoyu Ye, Hainan State Key Laboratory of Computer Architecture Institute of Computing Technology Chinese Academy of Sciences BenchCouncil [International Open Benchmarking Council Beijing Academy of Frontier Sciences and Technology University of Chinese Academy of Sciences Alibaba Baidu Tencent *** NetEase ByteDance Zhihu Lenovo Paypal Moqi Microsoft Research Asia Huawei *** CloudTa Intellifusion

Domain-specific software and hardware co-design is encouraging as it is much easier to achieve efficiency for fewer tasks. Agile domain-specific benchmarking speeds up the process as it provides not only relevant design inputs but also relevant metrics, and tools. Unfortunately, modern workloads like Big data, AI, and Internet services dwarf the traditional one in terms of code size, deployment scale, and execution path, and hence raise serious benchmarking challenges. This paper proposes an agile domain-specific benchmarking methodology. Together with seventeen industry partners, we identify ten important end-to-end application scenarios, among which sixteen representative AI tasks are distilled as the AI component benchmarks. We propose the permutations of essential AI and non-AI component benchmarks as end-to-end benchmarks. An end-to-end benchmark is a distillation of the essential attributes of an industry-scale application. We design and implement a highly extensible, configurable, and flexible benchmark framework, on the basis of which, we propose the guideline for building end-to-end benchmarks, and present the first end-to-end Internet service AI benchmark. The preliminary evaluation shows the value of our benchmark suite-AIBench against MLPerf and TailBench for hardware and software designers, micro-architectural researchers, and code developers. The specifications, source code, testbed, and results are publicly available from the web site http: //***/AIBench/***. Copyright © 2020, The Authors. All rights reserved.

关键词： Benchmarking

来源：评论

学校读者我要写书评

暂无评论

Provably Convergent Federated Trilevel Learning

arXiv

引用

arXiv 2023年

作者： Jiao, Yang Yang, Kai Wu, Tiancheng Jian, Chengtao Huang, Jianwei Department of Computer Science and Technology Tongji University China Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University China Shanghai Research Institute for Intelligent Autonomous Systems China School of Science and Engineering The Chinese University of Hong Kong Shenzhen China Shenzhen Institute of Artificial Intelligence and Robotics for Society China

Trilevel learning, also called trilevel optimization (TLO), has been recognized as a powerful modelling tool for hierarchical decision process and widely applied in many machine learning applications, such as robust neural architecture search, hyperparameter optimization, and domain adaptation. Tackling TLO problems has presented a great challenge due to their nested decision-making structure. In addition, existing works on TLO face the following key challenges: 1) they all focus on the non-distributed setting, which may lead to privacy breach;2) they do not offer any non-asymptotic convergence analysis which characterizes how fast an algorithm converges. To address the aforementioned challenges, this paper proposes an asynchronous federated trilevel optimization method to solve TLO problems. The proposed method utilizes µ-cuts to construct a hyper-polyhedral approximation for the TLO problem and solve it in an asynchronous manner. We demonstrate that the proposed µ-cuts are applicable to not only convex functions but also a wide range of non-convex functions that meet the µ-weakly convex assumption. Furthermore, we theoretically analyze the non-asymptotic convergence rate for the proposed method by showing its iteration complexity to obtain ϵ-stationary point is upper bounded by (Formula Presented). Extensive experiments on real-world datasets have been conducted to elucidate the superiority of the proposed method, e.g., it has a faster convergence rate with a maximum acceleration of approximately 80%. Copyright © 2023, The Authors. All rights reserved.

关键词： Decision making

来源：评论

学校读者我要写书评

暂无评论

Recent progress of integrated circuits and optoelectronic chips

引用

Science China(Information Sciences) 2021年第10期64卷 103-135页

作者： Yue HAO Shuiying XIANG Genquan HAN Jincheng ZHANG Xiaohua MA Zhangming ZHU Xingxing GUO Yahui ZHANG Yanan HAN Ziwei SONG Yan LIU Ling YANG Hong ZHOU Jiangyi SHI Wei ZHANG Min XU Weisheng ZHAO Biao PAN Yangqi HUANG Qi LIU Yimao CAI Jian ZHU Xin OU Tiangui YOU Huaqiang WU Bin GAO Zhiyong ZHANG Guoping GUO Yonghua CHEN Yong LIU Xiangfei CHEN Chunlai XUE Xingjun WANG Lixia ZHAO Xihua ZOU Lianshan YAN Ming LI School of Microelectronics Xidian University State Key Laboratory of Integrated Service Networks Xidian University National Integrated Circuit Innovation Center School of Microelectronics Fudan University School of Integrated Circuit Science and Engineering Beihang University China Center for Information Industry Development Frontier Institute of Chip and System Fudan University Institute of Microelectronics Peking University Nanjing Electronic Devices Institute State Key Laboratory of Functional Material for Informatics Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Institute of Microelectronics Tsinghua University Center for Carbon-based Electronics Department of Electronics Peking University CAS Key Laboratory of Quantum Information University of Science and Technology of China CAS Center for Excellence and Synergetic Innovation Center in Quantum Information and Quantum Physics University of Science and Technology of China Origin Quantum Computing Company Limited Key Laboratory of Flexible Electronics & Institute of Advanced Materials Nanjing Tech University State Key Laboratory of Electronic Thin Films and Integrated Devices School of Optoelectronic Science and EngineeringUniversity of Electronic Science and Technology of China National Laboratory of Solid State Microstructures & College of Engineering and Applied Sciences Nanjing University State Key Laboratory on Integrated Optoelectronics Institute of Semiconductors Chinese Academy of Sciences Center of Materials Science and Optoelectronics Engineering University of Chinese Academy of Sciences State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Engineering and Computer Science Peking University School of Electrical and Electronic Engineering Tiangong University Center for Information Photonics and Communications School of Information Science and TechnologySouthwest Jiaotong University School of Electronic Electrical and Communication Engineering

Integrated circuits(ICs) and optoelectronic chips are the foundation stones of the modern information society. The IC industry has been driven by the so-called "Moore's law" in the past 60 years, and now has entered the post Moore's law era. In this paper, we review the recent progress of ICs and optoelectronic chips. The research status, technical challenges and development trend of devices, chips and integrated technologies of typical IC and optoelectronic chips are focused on. The main contents include the development law of IC and optoelectronic chip technology, the IC design and processing technology, emerging memory and chip architecture, brain-like chip structure and its mechanism, heterogeneous integration, quantum chip technology, silicon photonics chip technology, integrated microwave photonic chip, and optoelectronic hybrid integrated chip.

关键词： integrated circuit semiconductor science and technology optoelectronic device and chip photonic integrated circuit wide bandgap semiconductors silicon photonics hybrid integration quantum chip integrated microwave photonic photonic neural computing

来源：评论

学校读者我要写书评

暂无评论

ShuffleInfer: Disaggregate LLM Inference for Mixed Downstream Workloads

引用

ACM Transactions on architecture and Code Optimization 1000年

作者： CunChen Hu HeYang Huang LiangLiang Xu XuSheng Chen Chenxi Wang Jiang Xu Shuang Chen Hao Feng Sa Wang Yungang Bao Ninghui Sun Yizhou Shan Institute of Computing Technology of the Chinese Academy of Sciences State Key Laboratory of Computer Architecture beijing Beijing China Institute of Computing Technology of the Chinese Academy of Sciences State Key Laboratory of Computer Architecture beijing Beijing China Xidian University Xian China Huawei Cloud Computing Technologies Co Ltd Guiyang China Institute of Computing Technology Chinese Academy of Sciences Beijing China State Key Laboratory of Computer Architecture Institute of Computing Technology of the Chinese Academy of Sciences beijing China State Key Laboratory of Computer Architecture Institute of Computing Technology Chinese Academy of Sciences Beijing China

Transformer-based large language model (LLM) inference serving is now the backbone of many cloud services. LLM inference consists of a prefill phase and a decode phase. However, existing LLM deployment practices often overlook the distinct characteristics of these phases, leading to significant interference. To mitigate interference, our insight is to carefully schedule and group inference requests based on their characteristics. We realize this idea in ShuffleInfer through three pillars. First, it partitions prompts into fixed-size chunks so that the accelerator always runs close to its computation-saturated limit. Second, it disaggregates prefill and decode instances so each can run independently. Finally, it uses a smart two-level scheduling algorithm augmented with predicted resource usage to avoid decode scheduling hotspots. Results show that ShuffleInfer improves time-to-first-token (TTFT), job completion time (JCT), and inference efficiency in turns of performance per dollar by a large margin, e.g., it uses 38% less resources all the while lowering average TTFT and average JCT by 97% and 47%, respectively.

关键词： LLM Serving Disaggregated Interference Schedule

来源：评论

学校读者我要写书评

暂无评论

Multi-modal Data based Semi-Supervised Learning for Vehicle Positioning

arXiv

引用

arXiv 2024年

作者： Huan, Ouwen Yang, Yang Luo, Tao Chen, Mingzhe Beijing Laboratory of Advanced Information Network Beijing University of Posts and Telecommunications Beijing100876 China Department of Electrical and Computer Engineering Institute for Data Science and Computing University of Miami Coral GablesFL33146 United States Beijing Key Laboratory of Network System Architecture and Convergence School of Information and Communication Engineering Beijing University of Posts and Telecommunications Beijing100876 China

In this paper, a multi-modal data based semisupervised learning (SSL) framework that jointly use channel state information (CSI) data and RGB images for vehicle positioning is designed. In particular, an outdoor positioning system where the vehicle locations are determined by a base station (BS) is considered. The BS equipped with several cameras can collect a large amount of unlabeled CSI data and a small number of labeled CSI data of vehicles, and the images taken by cameras. Although the collected images contain partial information of vehicles (i.e. azimuth angles of vehicles), the relationship between the unlabeled CSI data and its azimuth angle, and the distances between the BS and the vehicles captured by images are both unknown. Therefore, the images cannot be directly used as the labels of unlabeled CSI data to train a positioning model. To exploit unlabeled CSI data and images, a SSL framework that consists of a pretraining stage and a downstream training stage is proposed. In the pretraining stage, the azimuth angles obtained from the images are considered as the labels of unlabeled CSI data to pretrain the positioning model. In the downstream training stage, a small sized labeled dataset in which the accurate vehicle positions are considered as labels is used to retrain the model. Simulation results show that the proposed method can reduce the positioning error by up to 30% compared to a baseline where the model is not pretrained. Copyright © 2024, The Authors. All rights reserved.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

Visible Light Positioning Based on a Single Luminaire: A Novel Visual Odometry Assisted Algorithm

Visible Light Positioning Based on a Single Luminaire: A Nov...

引用

IEEE International Conference on Communications (ICC)

作者： Zhiyu Zhu Yang Yang Mingzhe Chen Caili Guo Yipeng Bai Beijing Key Laboratory of Network System Architecture and Convergence School of Information and Communication Engineering Beijing University of Posts and Telecommunications Beijing China Department of Electrical and Computer Engineering Institute for Data Science and Computing University of Miami Coral Gables FL USA Beijing Laboratory of Advanced Information Networks School of Information and Communication Engineering Beijing University of Posts and Telecommunications Beijing China

Visible light positioning (VLP) is a promising positioning technique, which, however, typically requires multiple luminaires to achieve accurate positioning. This paper proposes a novel visual odometry (VO) assisted visible light positioning algorithm (VO-VLP) in achieving positioning with only a single luminaire. In the considered model, a user equipped with a camera jointly uses geometric features in the captured images and coordinates information obtained via visible light communication (VLC) for positioning. The proposed VLP algorithm does not rely on any extra inertial measurement unit and relaxes the tilted angle limitation at the user. In particular, VO-VLP first uses the circle feature of a luminaire to obtain dual normal vectors of the luminaire. Then, the basic principle of VO is used to eliminate the wrong normal vector by exploiting the geometric features in two consecutive images captured when the user moves. Finally, the pose and location of the user are obtained by using an artificially marked point on the luminaire's contour. VO-VLP can achieve accurate positioning with only a single luminaire and a camera. Simulation results show that the proposed indoor positioning algorithm can achieve a 97th-percentile positioning accuracy of around 10 cm.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Lecture Notes in computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Preface

Lecture Notes in Computer Science (including subseries Lectu...

引用

Lecture Notes in computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 2013年 8299 LNCS卷 VI页

作者： Cohen, Albert Wu, Chenggang INRIA École Normale Supérieure Département d'Informatique 45 rue d'Ulm 75005 Paris France Chinese Academy of Sciences Institute of Computing Technology State Key Laboratory of Computer Architecture No. 6 Kexueyuan South Road Haidian District 100190 Beijing China

来源：评论

学校读者我要写书评

暂无评论

DSC: Deep scan context descriptor for large-scale place recognition

arXiv

引用

arXiv 2021年

作者： Cui, Jiafeng Huang, Tengfei Cai, Yingfeng Zhao, Junqiao Xiong, Lu Yu, Zhuoping School of Automobile Tongji University Shanghai China Department of Computer Science and Technology School of Electronics and Information Engineering Tongji University Shanghai China The Key Laboratory of Embedded System and Service Computing Ministry of Education Tongji University Shanghai China Institute of Intelligent Vehicles Tongji University Shanghai China

LiDAR-based place recognition is an essential and challenging task both in loop closure detection and global relocalization. We propose Deep Scan Context (DSC), a general and discriminative global descriptor that captures the relationship among segments of a point cloud. Unlike previous methods that utilize either semantics or a sequence of adjacent point clouds for better place recognition, we only use raw point clouds to get competitive results. Concretely, we first segment the point cloud egocentrically to acquire centroids and eigenvalues of the segments. Then, we introduce a graph neural network to aggregate these features into an embedding representation. Extensive experiments conducted on the KITTI dataset show that DSC is robust to scene variants and outperforms existing methods. © 2021, CC BY.

关键词： Eigenvalues and eigenfunctions

来源：评论

学校读者我要写书评

暂无评论

Large Scale Model Enabled Semantic Communications Based on Robust Knowledge Distillation

Large Scale Model Enabled Semantic Communications Based on R...

引用

IEEE Conference on Global Communications (GLOBECOM)

作者： Kuiyuan Ding Fangfang Liu Yang Yang Mingzhe Chen Caili Guo Beijing Key Laboratory of Network System Architecture and Convergence School of Information and Communication Engineering Beijing University of Posts and Telecommunications Beijing China Department of Electrical and Computer Engineering and the Institute for Data Science and Computing University of Miami Coral Gables FL USA Beijing Laboratory of Advanced Information Networks School of Information and Communication Engineering Beijing University of Posts and Telecommunications Beijing China

ISBN: (数字)9798350351255

ISBN: (纸本)9798350351262

Large scale artificial intelligence (AI) models possess excellent capabilities in semantic representation and understanding, making them particularly well-suited for semantic encoding and decoding. However, the substantial scale of these AI models imposes unacceptable computational resources and communication delays. To address this issue, we propose a semantic communication scheme based on robust knowledge distillation (RKD-SC) for large scale model enabled semantic communications. In the considered system, a transmitter extracts the features of the source image for robust transmission and accurate image classification at the receiver. To effectively utilize the superior capability of large scale model while make the cost affordable, we first transfer knowledge from a large scale model to a smaller scale model to serve as the semantic encoder. Then, to enhance the robustness of the system against channel noise, we propose a channel-aware autoencoder (CAA) based on the Transformer architecture. Experimental results show that the encoder of proposed RKD-SC system can achieve over 93.3% of the performance of a large scale model while compressing 96.67% number of parameters. Code: https://***/echojayne/RKD-SC.

关键词： Transmitters Computational modeling Noise Autoencoders Receivers Semantic communication Transformers Feature extraction Robustness Decoding

来源：评论

学校读者我要写书评

暂无评论

CSformer: Bridging Convolution and Transformer for Compressive Sensing

arXiv

引用

arXiv 2021年

作者： Ye, Dongjie Ni, Zhangkai Wang, Hanli Zhang, Jian Wang, Shiqi Kwong, Sam The Department of Computer Science City University of Hong Kong 999077 Hong Kong The Department of Computer Science & Technology Key Laboratory of Embedded System and Service Computing Ministry of Education Shanghai Institute of Intelligent Science and Technology Tongji University Shanghai200092 China The School of Electronic and Computer Engineering Peking University Shenzhen Graduate School Shenzhen518055 China The Peng Cheng Laboratory Shenzhen518052 China The City University of Hong Kong Shenzhen Research Institute Shenzhen518057 China

Convolution neural networks (CNNs) have succeeded in compressive image sensing. However, due to the inductive bias of locality and weight sharing, the convolution operations demonstrate the intrinsic limitations in modeling the long-range dependency. Transformer, designed initially as a sequence-to-sequence model, excels at capturing global contexts due to the self-attention-based architectures even though it may be equipped with limited localization abilities. This paper proposes CSformer, a hybrid framework that integrates the advantages of leveraging both detailed spatial information from CNN and the global context provided by transformer for enhanced representation learning. The proposed approach is an end-to-end compressive image sensing method, composed of adaptive sampling and recovery. In the sampling module, images are measured block-by-block by the learned sampling matrix. In the reconstruction stage, the measurement is projected into dual stems. One is the CNN stem for modeling the neighborhood relationships by convolution, and the other is the transformer stem for adopting global self-attention mechanism. The dual branches structure is concurrent, and the local features and global representations are fused under different resolutions to maximize the complementary of features. Furthermore, we explore a progressive strategy and window-based transformer block to reduce the parameter and computational complexity. The experimental results demonstrate the effectiveness of the dedicated transformer-based architecture for compressive sensing, which achieves superior performance compared to state-of-the-art methods on different datasets. Copyright © 2021, The Authors. All rights reserved.

关键词： Compressed sensing

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：