检索结果-内蒙古大学图书馆

Research on performance optimization of virtual data space across WAN

Frontiers of computer science 2024年第6期18卷 167-187页

作者： Jiantong HUO Zhisheng HUO Limin XIAO Zhenxue HE State Key Laboratory of Software Development Environment Beihang UniversityBeijing 100191China School of Computer Science and Engineering Beihang UniversityBeijing 100191China High Performance Computing Center Beihang UniversityBeijing 100191China Hebei Key Laboratory of Agricultural Big Data Hebei Agricultural UniversityBaoding 071001China

For the high-performance computing in a WAN environment,the geographical locations of national supercomputing centers are scattered and the network topology is complex,so it is difficult to form a unified view of *** aggregate the widely dispersed storage resources of national supercomputing centers in China,we have previously proposed a global virtual data space named GVDS in the project of“high performance computing Virtual Data Space”,a part of the National Key Research and Development Program of *** GVDS enables large-scale applications of the high-performance computing to run efficiently across ***,the applications running on the GVDS are often data-intensive,requiring large amounts of data from multiple supercomputing centers across *** this regard,the GVDS suffers from performance bottlenecks in data migration and access across *** solve the above-mentioned problem,this paper proposes a performance optimization framework of GVDS including the multitask-oriented data migration method and the request access-aware IO proxy resource allocation *** a WAN environment,the framework proposed in this paper can make an efficient migration decision based on the amount of migrated data and the number of multiple data sources,guaranteeing lower average migration latency when multiple data migration tasks are running in *** addition,it can ensure that the thread resource of the IO proxy node is fairly allocated among different types of requests(the IO proxy is a module of GVDS),so as to improve the application’s performance across *** experimental results show that the framework can effectively reduce the average data access delay of GVDS while improving the performance of the application greatly.

关键词： storage aggregation across WANs large-scale applications GVDS data migration allocation of IO proxy resource

来源：评论

学校读者我要写书评

暂无评论

BERD+: A Generic Sequential Recommendation Framework by Eliminating Unreliable Data with Item- and Attribute-level Signals

引用

ACM Transactions on Information Systems 2023年第2期42卷 1-33页

作者： Sun, Yatong Yang, Xiaochun Sun, Zhu Wang, Bin School of Computer Science and Engineering Northeastern University School of Computing Macquarie University China Australia Institute of High Performance Computing Centre for Frontier AI Research A*STAR Singapore School of Computer Science and Engineering Northeastern University China National Frontiers Science Center for Industrial Intelligence and Systems Optimization China Key Laboratory of Data Analytics and Optimization for Smart Industry [Northeastern University Ministry of Education China

A preliminary work of our study was published at IJCAI’21 [47], which is substantially extended in the following aspects: (1) In Section 1, we analyze the necessity of introducing item attributes for detecting unreliable instances, together with the problems and challenges that attributes may bring in. (2) In Section 2, we add discussions about the limitations of existing attribute-aware recommender systems (Section 2.2) and denoising methods (Section 2.3) in the context of detecting unreliable instances. (3) In Section 4.2, we further conduct an in-depth analysis at the attribute level to demonstrate the capability of attributes for rectifying instance loss and uncertainty, as well as the disturbance caused by attributes. (4) We generalize BERD to a generic framework BERD+ in Section 5.1, equipped with novel modules, i.e., HU-GCN (Section 5.2) and EPE (Section 5.4), which properly incorporate item attributes while reducing their disturbance for rectifying instance uncer tainty (Section 5.5) and loss (Section 5.6). The generic BERD+ can be flexibly plugged into existing SRSs for performance enhanced recommendation via eliminating unreliable data. (5) In Section 6.2, we apply our BERD+ framework to seven state-of-the-art SRSs on five real-world datasets to illustrate its superiority. (6) To avoid unfair comparison caused by item attributes, we build and compare with the baseline that combines the original BERD and an advanced attribute-aware recommender system, KSR [19]. (7) For more comprehensive comparison, in Section 6.2.2, we compare BRED+ with two state-of-the-art denoising approaches;in Section 6.2.3, to examine the efficacy of HU-GCN and EPE, we compare HU-GCN with various attribute embedding techniques, i.e., variants of graph neural networks, and compare EPE with different attribute fusing methods, i.e., adding, concatenation, and weighted sum. (8) In Section 6.2.4, a detailed ablation study is conducted to verify the effectiveness of each module of BERD+. (

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Vision Transformers with Hierarchical Attention

引用

Machine Intelligence Research 2024年第4期21卷 670-683页

作者： Yun Liu Yu-Huan Wu Guolei Sun Le Zhang Ajad Chhatkuli Luc Van Gool Institute for Infocomm Research(I2R) A*STARSingapore 138632Singapore Institute of High Performance Computing(IHPC) A*STARSingapore 138632Singapore Computer Vision Lab ETH ZürichZürich 8092Switzerland School of Information and Communication Engineering University of Electronic Science and Technology of China(UESTC)Chengdu 611731China

This paper tackles the high computational/space complexity associated with multi-head self-attention(MHSA)in vanilla vision *** this end,we propose hierarchical MHSA(H-MHSA),a novel approach that computes self-attention in a hierarchical ***,we first divide the input image into patches as commonly done,and each patch is viewed as a ***,the proposed H-MHSA learns token relationships within local patches,serving as local relationship ***,the small patches are merged into larger ones,and H-MHSA models the global dependencies for the small number of the merged *** last,the local and global attentive features are aggregated to obtain features with powerful representation *** we only calculate attention for a limited number of tokens at each step,the computational load is reduced ***,H-MHSA can efficiently model global relationships among tokens without sacrificing fine-grained *** the H-MHSA module incorporated,we build a family of hierarchical-attention-based transformer networks,namely *** demonstrate the superiority of HAT-Net in scene understanding,we conduct extensive experiments on fundamental vision tasks,including image classification,semantic segmentation,object detection and instance ***,HAT-Net provides a new perspective for vision *** and pretrained models are available at https://***/yun-liu/HAT-Net.

关键词： Vision transformer hierarchical attention global attention local attention scene understanding.

来源：评论

学校读者我要写书评

暂无评论

Optimizing Medical Imaging: high-performance Hardware for Image Processing, Machine Learning 24

Optimizing Medical Imaging: High-Performance Hardware for Im...

引用

7th International Conference on Digital Medicine and Image Processing, DMIP 2024

作者： Kalyan Chakravarthy, N.S. Balapanur, Mouli Chandra Pandian, Arun Nambi Pulikanti, Jyothi Prasad, D. Syed Masood, Jafar Ali Ibrahim High-Performance Computing Research Laboratory QIS College of Engineering and Technology Andhra Pradesh Ongole India Department of Internet of Things School of Computer Science and Engineering Vellore Institute of Technology Tamil Nadu Vellore India

ISBN: (纸本)9798400709586

The demand for high-performance hardware solutions for machine learning tasks is growing as medical imaging evolves. In this paper, we will focus on the latest hardware advanced technologies: GPUs, TPUs and FPGAs that can be used for image processing optimization in medical imaging such as brain tumor segmentation or chest X-ray classification. We test these devices on the BraTS and ChestX-ray14 datasets to show how running models quicker increases diagnostic performance. The study offers a comprehensive performance and energy efficiency evaluation of all the considered hardware solutions when training deep learning models for medical imaging, providing clinical guidelines to choose appropriate new-generation accelerators before engaging in expensive leasing contracts. Copyright © 2024 held by the owner/author(s). Publication rights licensed to ACM.

关键词： GPUS

来源：评论

学校读者我要写书评

暂无评论

Towards Harmless Rawlsian Fairness Regardless of Demographic Prior 38

Towards Harmless Rawlsian Fairness Regardless of Demographic...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Wang, Xuanqian Li, Jing Tsang, Ivor W. Ong, Yew-Soon School of Computer Science and Engineering Beihang University China Institute of High-Performance Computing Agency for Science Technology and Research Singapore Centre for Frontier AI Research Agency for Science Technology and Research Singapore College of Computing and Data Science Nanyang Technological University Singapore

Due to privacy and security concerns, recent advancements in group fairness advocate for model training regardless of demographic information. However, most methods still require prior knowledge of demographics. In this study, we explore the potential for achieving fairness without compromising its utility when no prior demographics are provided to the training set, namely harmless Rawlsian fairness. We ascertain that such a fairness requirement with no prior demographic information essential promotes training losses to exhibit a Dirac delta distribution. To this end, we propose a simple but effective method named VFair to minimize the variance of training losses inside the optimal set of empirical losses. This problem is then optimized by a tailored dynamic update approach that operates in both loss and gradient dimensions, directing the model towards relatively fairer solutions while preserving its intact utility. Our experimental findings indicate that regression tasks, which are relatively unexplored from literature, can achieve significant fairness improvement through VFair regardless of any prior, whereas classification tasks usually do not because of their quantized utility measurements. The implementation of our method is publicly available at https://***/wxqpxw/VFair. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Open-source Hardware: Different Approaches to Softcore implementation 6

Open-source Hardware: Different Approaches to Softcore imple...

引用

6th IEEE International Conference on Distributed computing, VLSI, Electrical Circuits and Robotics, DISCOVER 2022

作者： Hiremath, Sneha Chickerur, Satyadhyan Dandin, Joel Patil, Mihir Muddinkoppa, Bhoomi Adakoli, Shreerama Kle Technological University Centre for High Performance Computing School of Computer Science and Engineering Hubballi India

ISBN: (数字)9781665487160

ISBN: (纸本)9781665487160

Since the last few years, open-source hardware is preferred in the landscape of hardware development. Originally proprietary hardware was only developed by big multinational companies like AMD, Intel, and NVIDIA, which were expensive and general-purpose. Due to this, people started developing their own hardware. Open-source hardware has support from the opensource community, is developed at low cost, and has presently got a lot of support from developers and enthusiasts. Softcore is a microprocessor that can be implemented on FPGA using logic synthesis and can be fabricated as open-source hardware. This paper focuses on how various approaches to softcore implementation can be done on FPGA which would involve logic cells, CLBs, LUTs, RAM, etc. The soft cores which are built upon the RISC-V ISA, namely VexRISCV and RISC-V BOOM, as well as the softcores adhering to the POWER ISA, such as Microwatt and LibreSoC, are the main topics of discussion. These softcores are installed and simulated. To test the working of the ISA, some standard test cases and test suites are executed. Hardware implementation is done by porting the softcore onto a Xilinx Nexys A7-100T FPGA board. © 2022 IEEE.

关键词： Field programmable gate arrays (FPGA)

来源：评论

学校读者我要写书评

暂无评论

PancakeFS: A Write Efficiently and Read Optimized Filesystem 4

PancakeFS: A Write Efficiently and Read Optimized Filesystem

引用

4th International Conference on Consumer Electronics and computer engineering, ICCECE 2024

作者： Wang, Yongkang Liu, Yang Chen, Zhiguang School of Computer Science and Engineering Sun Yat-sen University Guangzhou China Institute for Quantum Information College of Computer National University of Defense Technology State Key Laboratory of High-Performance Computing Changsha China

ISBN: (纸本)9798350384062

With the development of emerging technologies such as cloud computing and large AI models (such as LLM), many applications have placed higher demands on the intensive read and write processing of massive data. Addressing the limitations of traditional file systems, we propose PancakeFS, an LSM-Tree-based file system. Leveraging the write-friendly characteristics of LSM-Tree, PancakeFS exhibits efficient write performance. We improve read performance through the design of an inode allocation algorithm based on subtree partitioning and optimization of the disk layout for hot data reconstruction. Our experimental results demonstrate that PancakeFS outperforms Ext4 and TableFS in terms of both read and write performance. © 2024 IEEE.

关键词： Data handling

来源：评论

学校读者我要写书评

暂无评论

Adaptive Occlusion Face Recognition Feature Fusion Network Based on Precise Guidance Mask Extraction

引用

Journal of Shanghai Jiaotong University (science) 2025年 1-12页

作者： Ye, Jihua Zhang, Lei Wang, Chao Tu, Qixuan Zheng, Rehua Huang, Huiyuan Jiang, Aiwen School of Computer and Information Engineering Jiangxi Normal University Nanchang330022 China Jiangxi Provincial Key Laboratory for High Performance Computing Jiangxi Normal University Nanchang330022 China State International Science & Technology Cooperation Base of Networked Supporting Software Jiangxi Normal University Nanchang330022 China

The uncertainty in the position and size of occluding objects greatly affects the extraction of identity features in facial recognition, which is a challenge that existing methods fail to effectively address. To tackle this issue, this paper introduces a novel feature fusion network for occluded facial recognition that adaptively handles arbitrary occlusions and accurately extracts occlusion information masks. The network effectively mitigates the impact of occlusions on facial identity features through a feature fusion module. The proposed method employs an occlusion classification module and an occlusion classification loss to precisely guide the extraction of accurate occlusion masks. Subsequently, an occlusion-aware hybrid attention module is introduced, which focuses on the relationship between facial features and occlusion masks across channel and spatial dimensions. This fusion strategy enables the extraction of facial identity features that are resilient to occlusions. Experimental results indicate that the proposed method holds a significant advantage in handling facial recognition tasks with arbitrary occlusions. On the LFW and MF1 synthetic occlusion datasets, the average accuracy rates reached 93.19% and 72.57%, respectively, showing improvements of 1.7 and 2.39 percent points over mainstream methods. © Shanghai Jiao Tong University 2025.

关键词： A feature fusion occluded facial recognition occlusion mask occlusion perception TP391.41

来源：评论

学校读者我要写书评

暂无评论

AsT: An Asymmetric-Sensitive Transformer for Osteonecrosis of the Femoral Head Detection 37

AsT: An Asymmetric-Sensitive Transformer for Osteonecrosis o...

引用

37th AAAI Conference on Artificial Intelligence, AAAI 2023

作者： Chen, Haoyang Liu, Shuai Lu, Feng Li, Wei Sheng, Bin Li, Mi Jin, Hai Zomaya, Albert Y. National Engineering Research Center for Big Data Technology and System Services Computing Technology and System Lab Cluster and Grid Computing Lab School of Computer Science and Technology Huazhong University of Science and Technology China Centre for Distributed and High Performance Computing School of Computer Science The University of Sydney Australia Department of Computer Science and Engineering Shanghai Jiao Tong University China Tongji Hospital Tongji Medical College Huazhong University of Science and Technology China

ISBN: (纸本)9781577358800

Early diagnosis of osteonecrosis of the femoral head (ONFH) can inhibit the progression and improve femoral head preservation. The radiograph difference between early ONFH and healthy ones is not apparent to the naked eye. It is also hard to produce a large dataset to train the classification model. In this paper, we propose Asymmetric-Sensitive Transformer (AsT) to capture the uneven development of the bilateral femoral head to enable robust ONFH detection. Our ONFH detection is realized using the self-attention mechanism to femoral head regions while conferring sensitivity to the uneven development by the attention-shared transformer. The real-world experiment studies show that AsT achieves the best performance of AUC 0.9313 in the early diagnosis of ONFH and can find out misdiagnosis cases firmly. Copyright © 2023, Association for the Advancement of Artificial Intelligence (***). All rights reserved.

关键词： Large dataset

来源：评论

学校读者我要写书评

暂无评论

Distributed Deep Learning with Data Parallelism for Diabetic Retinopathy Classification 1

引用

7th International Conference on Smart Trends in computing and Communications, SmartCom 2023

作者： Patil, Sanskruti S. Patil, Mahesh S. Chickerur, Satyadhyan Giraddi, Shantala Shahapur, Seetharam N. Hadalageri, Anup School of Computer Science and Engineering KLE Technological University Karnataka Hubballi580031 India Center for High Performance Computing KLE Technological University Karnataka Hubballi580031 India

ISBN: (数字)9789819907694

ISBN: (纸本)9789819907687

The large-scale model training is typically slower and necessitates high-performance workstations and Distributed Deep Learning (DDL). The DDL models trained on a massive volume of data can outperform single accelerators being used, i.e., the performance of deep learning (DL) models can be enhanced by using distributed and parallel deep learning methods. But the DDL models need to be redesigned and evaluated for particular hardware resources and applications, requiring the literature on various DDL model evaluations for assistance. Authors in the proposed work design and evaluate DDL with data parallelism for classifying Diabetic retinopathy (DR) images. A performance comparative study on the DDL strategy implemented on DenseNet is presented where the model is trained on multi-GPUs with counts from 1 to 8. The results show that the model takes less time when trained on multi-GPUs compared to single GPU. The detailed evaluation analysis of DDL with data parallelism is presented in this paper. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Graphics processing unit

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：