检索结果-内蒙古大学图书馆

ACM International conference Proceeding Series 2023年 855-855页

作者： Ziegeler, Sean Larsen, Matt Huebl, Axel Usher, Will Us Department of Defense Hpc Modernization Program/GDIT United States Luminary Cloud United States Lawrence Berkeley National Laboratory United States Intel Corporation United States

来源：评论

学校读者我要写书评

暂无评论

PMemTrace: Lightweight and Efficient Memory Access Monitoring for Persistent Memory 1

引用

22nd International conference on Algorithms and Architectures for Parallel Processing, ICA3PP 2022

作者： Zhang, Yushuqing Lu, Kai Wu, Zhenwei Zhang, Wenzhe College of Computer National University of Defense Technology Changsha China

ISBN: (数字)9783031226779

ISBN: (纸本)9783031226762

Persistent memory gains increasing popularity in recent years. Many persistent memory systems leverage memory access monitoring to achieve user-transparent crash consistency enforcement. However, traditional program analysis-oriented memory access monitoring mechanisms incur prohibitive performance overheads in persistent memory applications. Especially, for applications developed for systems with hybrid DRAM and persistent memory, existing memory tracking approaches cannot distinguish volatile and persistent memory writes, which leads to significant unnecessary runtime cost. This paper describes PMemTrace, a lightweight and efficient memory access monitoring approach for persistent memory. PMemTrace conducts best-effort pointer analysis to speculatively reduce as many redundant instrumentations to volatile writes as possible at compile time. Besides, PMemTrace enforces thread-local memory access permission control with the Memory Protection Keys (MPK) hardware primitive in recent Intel processors to track the mistakenly filtered persistent writes. The evaluation results show that PMemTrace substantially outperforms the state-of-the-art memory tracking systems. © 2023, Springer Nature Switzerland AG.

关键词： Dynamic random access storage

来源：评论

学校读者我要写书评

暂无评论

Multithreaded Parallelism for Heterogeneous Clusters of QPUs

Multithreaded Parallelism for Heterogeneous Clusters of QPUs

引用

ISC high performance 2024 Research Paper Proceedings (39th International conference)

作者： Philipp Seitz Manuel Geiger Christian B. Mendl Department of Computer Science Technical University of Munich TUM School of Computation Information and Technology Garching Germany Department of Computer Science Technical University of Munich TUM School of Computation Information and Technology TUM Institute for Advanced Study Garching Germany

In this work, we present MILQ, a quantum unrelated parallel machines scheduler and cutter. The setting of unrelated parallel machines considers independent hardware backends, each distinguished by differing setup and processing times. MILQ optimizes the total execution time of a batch of circuits scheduled on multiple quantum devices. It leverages state-of-the-art circuit-cutting techniques to fit circuits onto the devices and schedules them based on a mixed-integer linear program. Our results show a total improvement of up to 26 % compared to a baseline approach.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Towards Continual Federated Learning of Monocular Depth for Autonomous Vehicles 100

Towards Continual Federated Learning of Monocular Depth for ...

引用

100th IEEE Vehicular Technology conference, VTC 2024-Fall

作者： De S. Soares, Elton F. Campos, Carlos Alberto V. Ibm Research Brazil Research Lab RJ Rio de Janeiro20031-170 Brazil Federal University of the State of Rio de Janeiro Graduate Program in Informatics RJ Rio de Janeiro22290-255 Brazil

ISBN: (纸本)9798331517786

Recent investigations in computer vision for autonomous vehicles have focused on depth estimation from images, owing to its cost-efficiency and adaptability. Monocular depth estimation, using a single camera, is notable for its adaptability compared to binocular techniques that require two fixed cameras. Sophisticated methodologies employ self-supervised deep neural networks, while latest research proposes the use of federated learning to tackle crucial challenges for autonomous driving, such as data privacy, network usage, computation distribution, and connectivity robustness. Nevertheless, continual learning presents an additional challenge for both centralized and federated training, as updating models with novel datasets may induce the forgetting of previously acquired samples, thereby diminishing the model's versatility. Meanwhile, recent research indicates that experience replay strategies can be used to alleviate forgetting in autonomous driving use cases. We introduce ERFedSCDepth, an innovative approach amalgamating experience replay, federated learning, and deep self-supervision to allow the training of monocular depth estimators with high effectiveness and efficiency. Assessment using KITTI and DDAD datasets shows the efficacy of our approach, achieving enhanced continual learning performance over state-of-the-art baseline at standard depth metrics. © 2024 IEEE.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

Improved SwinUnet-based image segmentation method for charcoal slag of flame retardant material 5

Improved SwinUnet-based image segmentation method for charco...

引用

5th IEEE International conference on Civil Aviation Safety and Information Technology, ICCASIT 2023

作者： Yue, Wei Yang, Lei Zhao, Xiaofang Dongguan University of Technology School of Computer Science and Technology Dongguan China Dongguan University of Technology School of Electronic Engineering and Intelligence Dongguan China

ISBN: (纸本)9798350310603

The carbon residue formed from calcination of flame retardant materials is crucial to evaluate their performance and ensure quality control. The complex shape and texture variations of the charcoal dross make traditional segmentation methods difficult to perform accurately. We propose an enhanced SwinUNet method to achieve more precise segmentation of the charcoal residue region in flame retardant materials. Improvements involve adapting the original encoder to utilize a convolutional approach and adding a transformer to the skip connection for processing the feature map. Annotate images are fed into an encoder based on a convolutional neural network to extract features, which are then up-sampled in a Swin Transformer-based decoder and connected with shallow feature maps of different scales produced during the encoding phase. Finally, the improved SwinUNet method is compared to several classical image segmentation methods. The results demonstrate significant outperformance of the remaining methods with about a 2% improvement over SwinUnet on Dice. These findings offer an effective tool and guidance for evaluating the performance of flame retardant materials. © 2023 IEEE.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

Line-Start Permanent Magnet Machines for Low-Power Applications

Line-Start Permanent Magnet Machines for Low-Power Applicati...

引用

2024 International conference on Electrical Machines, ICEM 2024

作者： Barta, Jan Toman, Marek Losak, Petr Aarniovuori, Lassi Vitek, Ondrej Brno University of Technology Department of Power Electrical and Electronic Engineering Brno Czech Republic Institute of Automation and Computer Science Brno University of Technology Brno Czech Republic Electrical Engineering LUT University Lahti Finland

ISBN: (纸本)9798350370607

The improvement of efficiency is one of the main areas of research in electrical machines today. Three-phase induction machines are widely used in industry, motivating researchers to explore methods to increase their efficiency or replace them with more efficient alternatives. A line-start synchronous machine is a promising machine design for achieving high-efficiency classes even for low-power applications. The main contribution of this paper is to show the achievable performance of the LSPMSM by simply replacing the rotor of an induction machine. In this work, two different geometries suitable for low-power machines are optimized and studied in terms of electromagnetic parameters. Finally, the results of the start-up simulations for both machine designs are presented. The presented method shows that by replacing the induction machine rotor with a line-start permanent magnet rotor, it is possible to significantly enhance machine efficiency, even for low-power applications. © 2024 IEEE.

关键词： Permanent magnets

来源：评论

学校读者我要写书评

暂无评论

Learned Reverse ISP with Soft Supervision 17th

Learned Reverse ISP with Soft Supervision

引用

17th European conference on computer Vision, ECCV 2022

作者： Zou, Beiji Zhang, Yue School of Computer Science and Engineering Central South University Changsha410083 China Hunan Engineering Research Center of Machine Vision and Intelligent Medicine Changsha410083 China

ISBN: (纸本)9783031250620

RAW image serves as the foundation for camera imaging, which resides at the very beginning of the pipeline that generates sRGB images. Unfortunately, owing to special considerations, the information-rich RAW images are forfeited by default in most existing applications. To regain the RAW image, some works attempt to restore RAW images from RGB images. They focus on designing handcrafted model-based methods or complicated networks, however, ignoring the special property of RAW image, i.e., high dynamic range. To make up for this deficiency, we introduce a novel soft supervision, derived from the high dynamic range. Specifically, we propose to soften the original ground-truth as a multivariate Gaussian distribution so that networks could learn much more information. Then, we introduce a soft supervision driven network (SSDNet), based on convolution and transformer, for effectively restoring RAW images from RGB images. Quantitative and qualitative results show the promising restoration performance of RGB-to-RAW. In particular, our method achieved fifth place in the S7 track of AIM Reversed ISP Challenge. The source code will be available at https://***/yuezhang98/Learned-Reverse-ISP-with-Soft-Supervision. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

CFS: Scaling Metadata Service for Distributed File System via Pruned Scope of Critical Sections 23

CFS: Scaling Metadata Service for Distributed File System vi...

引用

18th European conference on computer Systems (EuroSys)

作者： Wang, Yiduo Wu, Yufei Li, Cheng Zheng, Pengfei Cao, Biao Sun, Yan Zhou, Fei Xu, Yinlong Wang, Yao Xie, Guangjun Univ Sci & Technol China Hefei Peoples R China Baidu China Co Ltd Beijing Peoples R China Anhui Prov Key Lab High Performance Comp Hefei Peoples R China

ISBN: (纸本)9781450394871

There is a fundamental tension between metadata scalability and POSIX semantics within distributed file systems. The bottleneck lies in the coordination, mainly locking, used for ensuring strong metadata consistency, namely, atomicity and isolation. CFS is a scalable, fully POSIX-compliant distributed file system that eliminates the metadata management bottleneck via pruning the scope of critical sections for reduced locking overhead. First, CFS adopts a tiered metadata organization to scale file attributes and the remaining namespace hierarchies independently with appropriate partitioning and indexing methods, eliminating cross-shard distributed coordination. Second, it further scales up the single metadata shard performance by single-shard atomic primitives, shortening the metadata requests' lifespan and removing spurious conflicts. Third, CFS drops the metadata proxy layer but employs the light-weight, scalable client-side metadata resolving. CFS has been running in the production environment of Baidu AI Cloud for three years. Our evaluation with a 50-node cluster and microbenchmarks shows that CFS simultaneously improves the throughput of baselines like HopsFS and InfiniFS by 1.76-75.82x and 1.22-4.10x, and reduces their average latency by up to 91.71% and 54.54%, respectively. Under cases with higher contention and larger directories, CFS' throughput benefits expand by one order of magnitude. For three real-world workloads with data accesses, CFS introduces 1.62-2.55x end-to-end throughput speedups and 35.06-62.47% tail latency reductions over InfiniFS.

关键词： Distributed file system Metadata management

来源：评论

学校读者我要写书评

暂无评论

Deep Neural Network Enhanced Chip-on-board Hybrid-Packaged Photonic-Electrical Silicon Transceiver for high-speed Optical Interconnects

Deep Neural Network Enhanced Chip-on-board Hybrid-Packaged P...

引用

2024 conference on Lasers and Electro-Optics, CLEO 2024

作者： Qin, Jun Yang, Qipeng Han, Changhao Zhou, Yan Zhang, Yunhao Wu, Yichen Sun, Yu Lu, Junde Liu, Siming Li, Yueqin Sun, Jian Min, Miao Hu, Weiwei He, Zhixue Wang, Lei Yu, Shaohua Shu, Haowen Wang, Xingjun Key Laboratory of the Ministry of Education for Optoelectronic Measurement Technology and Instrument Beijing Information Science and Technology University Beijing100192 China State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Peking University Beijing100871 China Department of Electrical and Computer Engineering University of California Santa BarbaraCA93106 United States Peking University Yangtze Delta Institute of Optoelectronics Nantong226010 China Peng Cheng Laboratory Shenzhen518055 China State Key Laboratory of Information Photonics and Optical Communications BUPT Beijing100876 China

We explore implementing a multilevel deep neural network to enhance the performance of a 4-channel photonic-electrical hybrid-packaged silicon transceiver. Stable transmission and reception of 150 Gbps/λ PAM4 signals... 详细信息

ISBN: (纸本)9781957171395

关键词： Optical interconnects

来源：评论

学校读者我要写书评

暂无评论

QSR: Q-Learning Based Algorithm of Support Recovery for Sparse Signals 100

QSR: Q-Learning Based Algorithm of Support Recovery for Spar...

引用

100th IEEE Vehicular Technology conference, VTC 2024-Fall

作者： Zhou, Tianzhi Wang, Yibo Wu, Zhangyan Liang, Hengwei Qi, Yuan Qian, Rongrong Beijing University of Posts and Telecommunications School of Electronic Engineering Beijing China Beijing University of Posts and Telecommunications School of Artificial Intelligence Beijing China Nanjing University of Science and Technology School of Computer Science and Engineering Nanjing China Beijing University of Posts and Telecommunications School of Intelligent Engineering and Automation Beijing China

ISBN: (纸本)9798331517786

Within a number of compressed sensing (CS) problems, support recovery of sparse signals is the primary task that needs to be performed. In this paper, we propose a Q-learning based algorithm of support recovery (QSR) for sparse signals, which treats the atomic selection in traditional greedy CS algorithms as Markov Decision Process (MDP) in reinforcement learning (RL). We adopt parallel computing to address "the curse of dimensionality"in Q-learning, dividing the overall RL problem into multiple subproblems. The computational complexity can thus significantly decrease, which grows linearly with the increase of signal length. Experiments show that the proposed algorithm performs closely to other CS algorithms when dealing with high sparsity level signals and it has better performance when dealing with low sparsity level signals. © 2024 IEEE.

关键词： Markov processes

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：