检索结果-内蒙古大学图书馆

Accelerating CNN Training With Concurrent Execution of GPU and Processing-in-Memory

IEEE ACCESS 2024年 12卷 160190-160204页

作者： Choi, Jungwoo Lee, Hyuk-Jae Sohn, Kyomin Yu, Hak-Soo Rhee, Chae Eun Seoul Natl Univ Interuniv Semicond Res Ctr ISRC Dept Elect Engn & Comp Sci Seoul 08826 South Korea Samsung Elect Hwaseong Si 18448 South Korea Hanyang Univ Dept Elect Engn Seoul 04763 South Korea

Training of convolutional neural networks (CNN) consumes a lot of time and resources. While most previous works have focused on accelerating the convolutional (CONV) layer, the proportion of non-convolutional (non-CONV) layers, such as batch normalization, is gradually increasing during training. Non-CONV layers have low cache reuse and arithmetic intensity, thereby performance is limited by memory bandwidth. Processing-in-memory (PIM) can utilize wide memory bandwidth, making it suitable for acceleration of non-CONV layers. Therefore, it makes sense to perform the computationally complex CONV layer on the host and handle the memory bottleneck challenges of the non-CONV layer on the PIM. Further improved performance can be expected if they run simultaneously. However, memory access conflicts between the host and PIM are the biggest factors hindering performance improvement. Prior studies proposed bank partitioning to alleviate memory conflicts, but it is not effective because CNN training involves significant data sharing between CONV and non-CONV layers. In this paper, we propose a memory scheduling and CNN training flow for the pipelined execution of CONV layers on the host and non-CONV layers on PIM. First, instead of applying bank partitioning, the host and PIM exclusively access memory for a certain period to avoid the movement of shared data between host memory and PIM memory. The conditions for switching the memory access authority between the host and PIM are set per layer, taking into account memory access characteristics and the number of queued memory requests. Second, in the training flow, CONV and non-CONV layers are pipelined in units of output feature map channels. Specifically, for the backward pass, the non-CONV tasks of the feature map gradient calculation phase and the weight gradient update phase are rearranged so that they can be easily performed within CONV layers. Experimental results show that the proposed pipelined execution achieves an aver

关键词： Training Graphics processing units Convolutional neural networks Pipelines Electric breakdown Bandwidth Batch normalization Switches scheduling algorithms Random access memory Processing-in-memory convolutional neural networks neural network training GPU

来源：评论

学校读者我要写书评

暂无评论

Distributed Stable Multi-Source Dynamic Broadcasting for Wireless Multi-Hop Networks Under SINR-Based Adversarial Channel Jamming

引用

IEEE-ACM TRANSACTIONS ON NETWORKING 2024年第6期32卷 5356-5371页

作者： Tian, Xiang Zhang, Baoxian Li, Cheng Yu, Jiguo Qilu Univ Technol Shandong Acad SciMinist Educ Key Lab Comp Power Network & Informat Secur Shandong Comp Sci CtrNatl Supercomp Ctr Jinan Jinan 250353 Peoples R China Shandong Fundamental Res Ctr Comp Sci Shandong Prov Key Lab Ind Network & Informat Syst Jinan 250353 Peoples R China Univ Chinese Acad Sci Res Ctr Ubiquitous Sensor Networks Beijing 100049 Peoples R China Simon Fraser Univ Sch Engn Sci Burnaby BC V5A 1S6 Canada Mem Univ Fac Engn & Appl Sci St John NF A1B 3X5 Canada Univ Elect Sci & Technol China Sch Comp Sci & Engn Chengdu 611731 Peoples R China Qilu Univ Technol Big Data Inst Jinan 250353 Peoples R China

Disseminating continuous packet flows injected at multiple location-random source nodes to all network nodes, known as the multi-source dynamic global broadcast problem, is a fundamental building block for wireless multi-hop networks to run smoothly and efficiently. Previous studies on dynamic global broadcast all assume reliable communications. However, in realistic wireless networks, there exist unpredictable transmission failures caused by the randomized signal interference from uncorrelated wireless networks sharing the same spectrum or even malicious attackers. In this paper, by integrating the Signal-to-Interference-plus-Noise-Ratio (SINR) model, multi-channel communication mode, and randomized malicious channel jamming controlled by an adaptive adversary, we present an SINR-based adversarial channel jamming model to capture the unpredictable transmission failures in a wireless multi-hop network. We first propose a distributed Jamming-resilient Multi-source Static Broadcast (JMSB) algorithm based on random channel selection and message transmissions for multi-hop wireless networks under the above SINR-based adversarial channel jamming model. We then propose a distributed stable Jamming-resilient Multi-source Dynamic Broadcast (JMDB) algorithm which iterates JMSB repeatedly and efficiently in a two-stage manner. We derive the maximum supportable broadcast throughput of JMDB under the stability guarantee, i.e., the expected boundedness on the queue length of each network node and expected broadcast latency for each injected packet. Simulation results shows the stability and throughput efficiency of our proposed JMDB algorithm.

关键词： Heuristic algorithms Jamming Stability analysis Spread spectrum communication scheduling algorithms Dynamic scheduling Wireless networks Throughput Signal to noise ratio Reliability Multi-source dynamic broadcast SINR channel jamming distributed algorithms stability

来源：评论

学校读者我要写书评

暂无评论

An Efficient Sub-6 GHz Assisted Millimeter Wave V2X Communication Scheduler

引用

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY 2024年第6期73卷 8915-8927页

作者： He, Chenyuan Zhang, Zhouyu Wan, Yan Zhao, Lu Lu, Hongsheng Shimizu, Takayuki Jiangsu Univ Sch Automot & Traff Engn Zhenjiang 212013 Peoples R China Univ Texas Arlington Dept Elect Engn Arlington TX 76019 USA Toyota Motor North America R&D InfoTech Labs Mountain View CA 94043 USA

The evolution of the Intelligent Transportation System (ITS) has driven dramatic increase in data exchange on road. Sub-6 GHz assisted millimeter wave (mmWave) technology is a promising solution to address this challenge. The sub-6 GHz channel serves as the control plane to facilitate peer discovery of mmWave equipped vehicles, while the mmWave technology provides high throughput data transmission. Considering the huge amounts of data and the conflicts among mmWave links, a sub-6 GHz assisted mmWave scheduler is indispensable. In this article, we study the mmWave scheduling problem that captures Vehicle-to-Everything (V2X) communication features, i.e., data importance and data freshness. We formulate the scheduling problem into a novel mixed-integer sum-of-ratios problem by leveraging discretized mmWave transmission slots. For this nonlinear NP-hard problem, we derive an equivalent mixed-integer linear programming (MILP) problem with significantly reduced computational complexity. Extensive comparative simulation studies with the baseline brute-force algorithm, the greedy coloring algorithm, our previously developed Maximum Weight Matching-based Iterative Algorithm (MWMIA) and Parameterization-based Iterative Algorithm (PIA) are conducted to verify the correctness and effectiveness of our proposed algorithm.

关键词： Millimeter wave communication Vehicle-to-everything Schedules Delays Optimal scheduling Data communication scheduling algorithms Vehicular networks vehicle-to-everything millimeter wave communication scheduling

来源：评论

学校读者我要写书评

暂无评论

Free Space Optical Communication for Inter-Satellite Link: Architecture, Potentials and Trends

引用

IEEE COMMUNICATIONS MAGAZINE 2024年第3期62卷 110-116页

作者： Wang, Guanhua Yang, Fang Song, Jian Han, Zhu Tsinghua Univ Beijing Peoples R China Univ Houston Houston TX USA Kyung Hee Univ Seoul South Korea

The sixth-generation network is expected to achieve global coverage based on the space-air-ground integrated network, and the latest satellite network will play an important role in it. The introduction of inter-satellite links (ISLs) can significantly improve the throughput of the satellite network, and has recently received a lot of attention from both academia and industry. In this article, we illustrate the advantages of using the laser for ISLs due to its longer communication distance, higher data speed, and stronger security. Specifically, space-borne laser terminals - with the acquisition, pointing, and tracking mechanism, which realizes long-distance communication - are illustrated. Advanced modulation and multiplexing modes that make high communication rates possible are introduced. The security of ISLs ensured by the characteristics of both laser and the optical channel is also analyzed. Moreover, some open issues, such as advanced optical beam steering, routing and scheduling algorithm, and integrated sensing and communication are discussed to direct future research.

关键词： 6G mobile communication Laser modes Security Satellite communication Throughput Space-air-ground integrated networks Routing protocols Optical sensors Optical modulation Optical beams scheduling algorithms Optical fiber communication

来源：评论

学校读者我要写书评

暂无评论

Real-Time Analysis and Message Priority Assignment for TSN-CAN Gateway

引用

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 2024年第11期25卷 16133-16144页

作者： Wu, Wufei Huang, Huijuan Li, Wenhao Liu, Ruihua Xie, Yong Long, Saiqin Nanchang Univ Sch Informat Engn Nanchang 330031 Peoples R China Nanchang Univ Sch Math & Comp Sci Nanchang 330000 Peoples R China Nanjing Univ Posts & Telecommun Sch Comp Sci Nanjing 210000 Peoples R China Jinan Univ Coll Informat Sci & Technol Guangzhou 510632 Peoples R China

As automobiles continue to develop in the direction of intelligence and networking, the requirements for in-vehicle network bandwidth and deterministic time delay continue to increase. However, existing in-vehicle network standard protocols such as Controller Area Network (CAN) cannot meet the increasing bandwidth needs of in-vehicle networks, and Time Sensitive Networking (TSN) has emerged a research hot-spot for next-generation in-vehicle network standards. The next-generation in-vehicle network is developing towards a domain network architecture with TSN as the backbone and other conventional buses as branches. In this architecture, the TSN-CAN gateway is an important component that handles the data communication between the TSN domain and CAN. Due to the large difference in transmission rates between TSN and CAN, the TSN-CAN heterogeneous gateway suffers from congestion resulting in unguaranteed real-time transmission of messages across the gateway. To address this issue, a high response ratio priority scheduling algorithm (HRRP) for TSN-CAN gateways based on worst-case response time analysis theory is proposed in this paper. The algorithm assigns forwarding priority to CAN messages based on the value of their response ratio, and the experimental demonstrate show that the method can significantly improve the schedulability and reduce latency, improving the real-time performance of the system.

关键词： Logic gates Real-time systems Ethernet Bandwidth scheduling scheduling algorithms Automobiles In-vehicle network gateway time sensitive network scheduling worst-case response time high response ratio priority

来源：评论

学校读者我要写书评

暂无评论

Comparison of Decentralized and Centralized Update Paradigms for Distributed Remote Estimation

引用

IEEE-ACM TRANSACTIONS ON NETWORKING 2024年第2期32卷 1839-1853页

作者： Kang, Sunjung Eryilmaz, Atilla Joo, Changhee Ohio State Univ Dept Elect & Comp Engn Columbus OH 43210 USA Korea Univ Dept Comp Sci & Engn Seoul 02841 South Korea

In this work, we perform a comparative study of centralized and decentralized update strategies for the basic remote tracking problem of many distributed users/devices with randomly evolving states. Our goal is to reveal the impact of the fundamentally different tradeoffs that exist between information accuracy and communication cost under these two update paradigms. In one extreme, decentralized updates are triggered by distributed users/transmitters based on exact local state-information, but also at a higher cost due to the need for uncoordinated multi-user communication. In the other extreme, centralized updates are triggered by the common tracker/receiver based on estimated global state-information, but also at a lower cost due to the capability of coordinated multi-user communication. We use a generic superlinear function to model the communication cost with respect to the number of simultaneous updates for multiple sources. We characterize the conditions under which transmitter-driven decentralized update policies outperform their receiver-driven centralized counterparts for symmetric sources, and vice versa. Further, we extend the results to a scenario where system parameters are unknown and develop learning-based update policies that asymptotically achieve the minimum cost levels attained by the optimal policies.

关键词： Remote estimation (RE) asymptotic analysis scheduling algorithms Internet of Things

来源：评论

学校读者我要写书评

暂无评论

TA2LS: A Traffic-Aware Multipath Scheduler for Cost-Effective QoE in Dynamic HetNets

引用

IEEE TRANSACTIONS ON MOBILE COMPUTING 2024年第12期23卷 13603-13620页

作者： Zhang, Xiaoya Zhang, Yuyang Dong, Ping Yang, Dong Du, Xiaojiang Yu, Chengxiao Zhang, Hongke Beijing Jiaotong Univ Beijing 100082 Peoples R China Stevens Inst Technol Dept Elect & Comp Engn Hoboken NJ 07030 USA Peng Cheng Lab Shenzhen 518066 Peoples R China

Multipath transmission is a critical enabling technology to enhance QoE for edge users. The packet scheduler plays an irreplaceable role in overcoming heterogeneity and dynamicity in multipath transmission. However, current schedulers depend on an inaccurate delay estimation and lack systematic traffic intensity awareness, performing poorly in wireless heterogeneous networks (HetNets). In this paper, we propose a novel traffic-aware two-level packet scheduler (TA2LS) to address the problem and improve aggregated bandwidth while trading off delay. In particular, we design a multipath transmission state machine (MTSM) to perceive link traffic intensity. MTSM replaces network prediction algorithms by identifying the contribution of each link in multipath transmission in a cost-effective way. Further, we propose a scheduling mechanism based on a two-level optimal-path evaluation method (2LOSM) to adjust the packet scheduling policy adaptively. 2LOSM increases the priority of links with low traffic intensity during scheduling, improving aggregated bandwidth performance and reducing end-to-end delay. We have built a real-world 4G/5G/WiFi testbed and deployed 47 dynamic scenarios to evaluate TA2LS and other five schedulers. In 4G/5G/WiFi scenarios, TA2LS improves aggregated bandwidth by 10.32%-48.27% compared to the second-best scheduler and reduces end-to-end delay by 5.04%-39.98% under the premise of fewer or equivalent overheads.

关键词： Delays Bandwidth Real-time systems Dynamic scheduling scheduling algorithms Quality of experience Wireless communication Multipath transmission packet scheduler traffic-aware heterogeneous networks

来源：评论

学校读者我要写书评

暂无评论

Channel-Aware Rotating Priority Queue scheduling with Delay Deadline over Wireless Links 16

Channel-Aware Rotating Priority Queue Scheduling with Delay ...

引用

16th International Conference on Wireless Communications and Signal Processing, WCSP 2024

作者： Zhang, Yankai Wang, Wei Yang, Xinglin Wang, Yitu Zhang, Wenkai Zhang, Zhaoyang College of Information Science and Electronic Engineering Zhejiang University Hangzhou China School of Electrical and Information Engineering North Minzu University Yinchuan China

ISBN: (纸本)9798350390643

Conventional Rotating Priority Queues (RPQ) scheduling adopted in Time-Sensitive Networks (TSN) fails to achieve desired performance over time-varying wireless channel. In this paper, we propose a channel-aware RPQ scheduling algorithm to meet explicit delay deadlines in wireless context. Specifically, we first construct an improved channel-aware RPQ framework. Different to the classic RPQ scheduler, we allocate the transmission opportunity to various devices before adopting RPQ scheduling for each device. For the scheduling of device transmission time, we establish a weighted queue for each device by learning their effects to the violation of delay deadlines, and propose a low-complexity scheduling algorithm by stabilizing the virtual outage queue via Lyapunov optimization. As the mapping to the weighted queue is non-smooth, we enhance the soft threshold function by employing momentum to accelerate the convergence rate. Simulation results verify the performance improvements of the proposed scheduling algorithm, © 2024 IEEE.

关键词： scheduling algorithms

来源：评论

学校读者我要写书评

暂无评论

Design and Optimization of Power System Day-Ahead scheduling Based on Kantorovich Algorithm 6

Design and Optimization of Power System Day-Ahead Scheduling...

引用

6th International Conference on Energy Systems and Electrical Power, ICESEP 2024

作者： Ai, Xin CO. LTD. Beijing100120 China

ISBN: (纸本)9798350353563

In this study, we enhance the k-means clustering algorithm for the initial classification of raw scenes. This process is complemented by a synchronized back generation reduction algorithm, which leverages Kantorovich probability distance to streamline scene processing, thereby reducing the computational demands of scene analysis. We have developed a power system day-ahead optimal dispatch model including wind power, photovoltaic (PV), thermal power, and pumped storage, which is based on refined scenario analyses and investigates the role of pumped storage units in improving the rate of wind and solar energy consumption. The effectiveness of our proposed model and methodology is demonstrated through simulation analysis using real data from a provincial grid. © 2024 IEEE.

关键词： scheduling algorithms

来源：评论

学校读者我要写书评

暂无评论

An enhanced exponential distribution optimizer for solving task scheduling in cloud computing systems 4

An enhanced exponential distribution optimizer for solving t...

引用

4th International Conference on Advanced algorithms and Neural Networks, AANN 2024

作者： Wu, Fengbin Li, Shaobo Wu, Libang Xiong, Xuan Xie, Rongxiang State Key Laboratory of Public Big Data Guizhou University China Guizhou Institute of Technology China College of Computer Science and Technology Guizhou University China

ISBN: (纸本)9781510686106

Cloud computing has become integral to information technology, offering a flexible and scalable way to access and utilize computing resources. In cloud computing, task scheduling policies can affect the resource usage efficiency of the underlying system. Hence, allocating user-input tasks to appropriate computing resources is an essential issue in cloud task scheduling. For this, many meta-heuristic algorithms have been introduced. This paper addresses the problem of cloud task scheduling using the exponential distribution optimizer (EDO) for the first time. Meanwhile, an enhanced EDO variant, called EEDO, is proposed to enhance the optimal solution search ability further. Specifically, a chaotic oppositional learning strategy is proposed to improve population diversity. Then, a guided solution weight adaptation technique is developed to augment the accuracy. Experimental findings indicate that the proposed EEDO makes better use of system resources for different tasks. Hence, EEDO can be considered as a promising scheduling method. © 2024 SPIE.

关键词： scheduling algorithms

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：