检索结果-内蒙古大学图书馆

Optimizing Winograd-Based Fast Convolution Algorithm on Phytium Multi-Core CPUs

Jisuanji Yanjiu yu Fazhan/Computer Research and Development 2020年第6期57卷 1140-1151页

作者： Wang, Qinglin Li, Dongsheng Mei, Songzhu Lai, Zhiquan Dou, Yong Science and Technology on Parallel and Distributed Processing Laboratory National University of Defense Technology Changsha410073 China College of Computer National University of Defense Technology Changsha410073 China

Convolutional neural networks (CNNs) have been extensively used in artificial intelligence fields such as computer vision and natural language processing. Winograd-based fast convolution algorithms can effectively reduce the computational complexity of convolution operations in CNNs so that they have attracted great attention. With the application of Phytium multi-core CPUs independently developed by the National University of Defense technology in artificial intelligence fields, there is strong demand of high-performance convolution primitives for Phytium multi-core CPUs. This paper proposes a new high-performance parallel Winograd-based fast convolution algorithm after studying architecture characteristics of Phytium multi-core CPUs and computing characteristics of Winograd-based fast convolution algorithms. The new parallel algorithm does not rely on general matrix multiplication routines, and consists of four stages: kernels transformation, input feature maps transformation, element-wise multiplication, and output feature maps inverse transformation. The data movements in all four stages have been collaboratively optimized to improve memory access performance of the algorithm. The custom data layouts, multi-level parallel data transformation algorithms and multi-level parallel matrix multiplication algorithm have also been proposed to support the optimization above efficiently. The algorithm is tested on two Phytium multi-core CPUs. Compared with Winograd-based fast convolution implementations in ARM Computer Library (ACL) and NNPACK, the algorithm can achieve speedup of 1.05~16.11 times and 1.66~16.90 times, respectively. The application of the algorithm in the open source framework Mxnet improves the forward-propagation performance of the VGG16 network by 3.01~6.79 times. © 2020, science Press. All right reserved.

关键词： Program processors

来源：评论

学校读者我要写书评

暂无评论

Research on Pin-by-Pin Calculation Method of Rectangular Mesh Based on Quasi-Diffusion Theory

SSRN

引用

SSRN 2024年

作者： Zhuang, Kun Yongzhan, Wang Yan, Jiangtao Zou, Hang Deng, Lina Wang, Yingzhen Wang, Sipeng Zhang, Qian Department of Nuclear Science and Technology Nanjing University of Aeronautics and Astronautics Jiangsu Nanjing211106 China China Nuclear Power Operation Technology Corporation LTD. Research Institute of Nuclear Power Operation Hubei Wuhan430223 China Science and Technology on Parallel and Distribution Processing Laboratory National University of Defense Technology Hunan Changsha410073 China Laboratory for Advanced Nuclear Energy Theory and Applications Zhejiang Institute of Modern Physics Department of Physics Zhejiang University Zhejiang Hangzhou310027 China

With the development of nuclear energy techniques, pin-by-pin calculation using the SP3 method based on pin cell homogenization attracts researchers. This study proposed a pin-by-pin calculation method of rectangular mesh based on quasi-diffusion (QD) theory. The QD equation and boundary condition were derived from a transport equation similar to the P1 equation derivation. The Eddington factor was treated as a few-group homogenized parameter and calculated based on a deterministic code ALPHA. The numerical method of QD equation was developed based on nodal expansion method of nonlinear iteration. Some benchmarks were used to verify the Eddington factor and the numerical method of QD. 4 colorset models with different neutron anisotropy levels were analyzed by pin-by-pin-QD method. Pin-by-pin-QD can reduce keff error by about 600 pcm and agree better with reference results. Additionally, it has better accuracy and similar efficiency as diffusion calculation and is suitable for problems with strong anisotropy. © 2024, The Authors. All rights reserved.

关键词： Numerical methods

来源：评论

学校读者我要写书评

暂无评论

Big data collaborative artificial intelligence and high-performance computing to drive physical design of fusion

Big data collaborative artificial intelligence and high-perf...

引用

Cloud Computing, Big Data Applications and Software Engineering (CBASE), International Conference on

作者： Pan Liu Gongjie Liu Jie Liu Chencheng Liu Wenbin Xiong Zhangcun Tang Yinzhao Zhou Yonggang Wang Zhouming Yuan Tianfu Innovation Energy Establishment Chengdu China Beijng PARATERA Technology Corp. Ltd Beijing China Science and Technology on Parallel and Distributed Processing Laboratory National University of Defense Technology Changsha China Laboratory of Software Engineering for Complex Systems National University of Defense Technology Changsha China Central South University of Forestry and Technology Changsha China Nuclear and Radiation Safety Center MEE Beijing China University of Electronic Science and Technology of China Chengdu China

Controlled thermonuclear fusion has always been a dream pursued by mankind. However, the physical processes of controlled thermonuclear fusion are complex, requiring numerical simulations with high performance computing, and the amount of data generated by the physical processes on spatial, temporal and temperature scales is too large to be captured, managed, processed and collated in a reasonable time frame by mainstream software tools to achieve more aggressive fusion physical design. The data are too large to be captured, managed, processed, and collated into more aggressive targets for fusion physical design in a reasonable time by mainstream software tools. At the same time, the failure of fusion ignition can be caused by the distortion of various key physical quantities, and only by decomposing the process step by step and clarifying the changes of key physical quantities in the fusion physics process, can an effective mechanism be formed to prevent the distortion of key physical quantities from causing ignition failure in experimental physics. Big data in collaboration with artificial intelligence and high performance computing to drive the physical design of fusion is a novel avenue. By data acquisition with and pre-processing, this allows the creation of small sample libraries and deep learning. With supervised learning function convergence, incorporating solid/fluid computational methods, further network layering and cell expansion, the parameters of the physical model conforming to Lawson’s criterion will become experimental physical parameters, and we find the new approach in Data capacity, Model combination approach, Type of material, Calculation speed, Optimisation of design iteration times, etc. are superior to the traditional approach.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Vision-Language Navigation: A Survey and Taxonomy

arXiv

引用

arXiv 2021年

作者： Wu, Wansen Chang, Tao Li, Xinmeng Yin, Quanjun Zhu, Jiancheng The College of Systems Engineering National University of Defense Technology Hunan Changsha410073 China National Key Laboratory of Parallel and Distributed Processing College of Computer Science and Technology National University of Defence Technology Changsha410073 China

Vision-Language Navigation (VLN) tasks require an agent to follow human language instructions to navigate in previously unseen environments. This challenging field involving problems in natural language processing, computer vision, robotics, etc., has spawn many excellent works focusing on various VLN tasks. This paper provides a comprehensive survey and an insightful taxonomy of these tasks based on the different characteristics of language instructions in these tasks. Depending on whether the navigation instructions are given for once or multiple times, this paper divides the tasks into two categories, i.e., single-turn and multi-turn tasks. For single-turn tasks, we further subdivide them into goal-oriented and route-oriented based on whether the instructions designate a single goal location or specify a sequence of multiple locations. For multi-turn tasks, we subdivide them into passive and interactive tasks based on whether the agent is allowed to question the instruction or not. These tasks require different capabilities of the agent and entail various model designs. We identify progress made on the tasks and look into the limitations of existing VLN models and task settings. Finally, we discuss several open issues of VLN and point out some opportunities in the future, i.e., incorporating knowledge with VLN models and implementing them in the real physical world. © 2021, CC BY.

关键词： Taxonomies

来源：评论

学校读者我要写书评

暂无评论

PCGraph: Accelerating GNN Inference on Large Graphs via Partition Caching

PCGraph: Accelerating GNN Inference on Large Graphs via Part...

引用

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

作者： Lizhi Zhang Zhiquan Lai Yu Tang Dongsheng Li Feng Liu Xiaochun Luo National Laboratory for Parallel and Distributed Processing (PDL) Computer College National University of Defence Technology Changsha China PLA News Media Center Beijing China

Graph neural networks (GNNs) have been emerging as powerful learning tools for unstructured data and successfully applied to many graph-based application domains. Sampling-based GNN inference is commonly adopted in existing graph learning frameworks to handle large-scale graphs. However, this approach is restricted by the problems of redundant vertex embedding computation in GPU and inefficient loading of vertex features from CPU to GPU. In this paper, we propose PCGraph, a system that supports adaptive GNN inference and feature partition caching. PCGraph significantly reduces the vertex embedding computation time by adaptive GNN inference technology, which selects the optimal inference algorithm and minimizes vertex embedding computation. PCGraph also reduces the redundant data transfer between the CPU and GPU by partition the target vertices and caching their corresponding partitions in turn. We evaluate PCGraph against two state-of-the-art industrial GNN frameworks, i.e., PyG and DGL, on a diverse array of benchmarks. Experimental results show that PCGraph reduces up to 99% vertex embedding computation and 98.5% data loading time, and achieves up to 360× performance speedup over the state-of-the-art baselines.

关键词： Adaptive systems Loading Graphics processing units Benchmark testing Data transfer Inference algorithms Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Excitement Surfeited Turns to Errors: Deep Learning Testing Framework Based on Excitable Neurons

arXiv

引用

arXiv 2022年

作者： Jin, Haibo Chen, Ruoxi Zheng, Haibin Chen, Jinyin Cheng, Yao Yu, Yue Liu, Xianglong College of Information Engineering Zhejiang University of Technology Hangzhou China Institute of Cyberspace Security Zhejiang University of Technology Hangzhou China Huawei International Singapore National Laboratory for Parallel and Distributed Processing College of Computer National University of Defense Technology Changsha China State Key Laboratory of Software Development Environment Beihang University Beijing China

Despite impressive capabilities and outstanding performance, deep neural networks (DNNs) have captured increasing public concern about their security problems, due to their frequently occurred erroneous behaviors. Therefore, it is necessary to conduct a systematical testing for DNNs before they are deployed to real-world applications. Existing testing methods have provided fine-grained metrics based on neuron coverage and proposed various approaches to improve such metrics. However, it has been gradually realized that a higher neuron coverage does not necessarily represent better capabilities in identifying defects that lead to errors. Besides, coverage-guided methods cannot hunt errors due to faulty training procedure. So the robustness improvement of DNNs via retraining by these testing examples are unsatisfactory. To address this challenge, we introduce the concept of excitable neurons based on Shapley value and design a novel white-box testing framework for DNNs, namely DeepSensor. It is motivated by our observation that neurons with larger responsibility towards model loss changes due to small perturbations are more likely related to incorrect corner cases due to potential defects. By maximizing the number of excitable neurons concerning various wrong behaviors of models, DeepSensor can generate testing examples that effectively trigger more errors due to adversarial inputs, polluted data and incomplete training. Extensive experiments implemented on both image classification models and speaker recognition models have demonstrated the superiority of DeepSensor. Compared with the state-of-the-art testing approaches, DeepSensor can find more test errors due to adversarial inputs (∼ ×1.2), polluted data (∼ ×5) and incompletely-trained DNNs (∼ ×1.3). Additionally, it can help DNNs build larger l2-norm robustness bound (∼ ×3) via retraining according to CLEVER's certification. We further provide interpretable proofs for effectiveness of DeepSensor via excitable neuro

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

NUMA-aware FFT-based convolution on ARMv8 many-core CPUs

arXiv

引用

arXiv 2021年

作者： Huang, Xiandong Wang, Qinglin Lu, Shuyu Hao, Ruochen Mei, Songzhu Liu, Jie Science and Technology on Parallel and Distributed Processing Laboratory National University of Defense Technology Changsha410073 China School of Computer Science National University of Defense Technology Changsha410073 China Department of Biomedical Informatics University of Pittsburgh Pittsburgh United States

Convolutional Neural Networks (CNNs), one of the most representative algorithms of deep learning, are widely used in various artificial intelligence applications. Convolution operations often take most of the computational overhead of CNNs. The FFT-based algorithm can improve the efficiency of convolution by reducing its algorithm complexity, there are a lot of works about the high-performance implementation of FFT-based convolution on many-core CPUs. However, there is no optimization for the non-uniform memory access (NUMA) characteristics in manycore CPUs. In this paper, we present a NUMA-aware FFT-based convolution implementation on ARMv8 many-core CPUs with NUMA architectures. The implementation can reduce a number of remote memory access through the data reordering of FFT transformations and the three-level parallelization of the complex matrix multiplication. The experiment results on a ARMv8 many-core CPU with NUMA architectures demonstrate that our NUMA-aware implementation has much better performance than the state-of-the-art work in most cases. Copyright © 2021, The Authors. All rights reserved.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Outlier detection ensemble with embedded feature delection

arXiv

引用

arXiv 2020年

作者： Cheng, Li Wang, Yijie Liu, Xinwang Li, Bin Science and Technology on Parallel and Distributed Processing Laboratory College of Computer National University of Defense Technology Changsha China

Feature selection places an important role in improving the performance of outlier detection, especially for noisy data. Existing methods usually perform feature selection and outlier scoring separately, which would select feature subsets that may not optimally serve for outlier detection, leading to unsatisfying performance. In this paper, we propose an outlier detection ensemble framework with embedded feature selection (ODEFS), to address this issue. Specifically, for each random sub-sampling based learning component, ODEFS unifies feature selection and outlier detection into a pairwise ranking formulation to learn feature subsets that are tailored for the outlier detection method. Moreover, we adopt the thresholded self-paced learning to simultaneously optimize feature selection and example selection, which is helpful to improve the reliability of the training set. After that, we design an alternate algorithm with proved convergence to solve the resultant optimization problem. In addition, we analyze the generalization error bound of the proposed framework, which provides theoretical guarantee on the method and insightful practical guidance. Comprehensive experimental results on 12 real-world datasets from diverse domains validate the superiority of the proposed ODEFS. Copyright © 2020, The Authors. All rights reserved.

关键词： Feature extraction

来源：评论

学校读者我要写书评

暂无评论

LTKT: Knowledge Tracing Based on Positive and Negative Learning Transfers

SSRN

引用

SSRN 2023年

作者： Xu, Jia Tang, Rongrong Lv, Pin Yu, Minghe Yu, Ge Chen, Enhong School of Computer Electronics and Information Guangxi University GuangXi Nanning530004 China Key Laboratory of Parallel Distributed and Intelligent Computing Education Department of Guangxi Zhuang Autonomous Region Guangxi530004 China Northeastern University Shenyang110819 China University of Science and Technology of China Hefei230088 China

Knowledge Tracing (KT) is a critical but challenging problem for many educational applications. As an essential part of educational psychology, the propagated influence among pedagogical concepts (i.e., learning transfer) is important for optimizing KT tasks. However, existing KT methods only consider the positive learning transfer and neglect the negative learning transfer. To this end, in this paper, we propose a novel deep knowledge tracing model, called positive and negative Learning Transfers based Knowledge Tracing (LTKT). To the best of our knowledge, LTKT makes the first attempt to concurrently utilize the positive and negative learning transfer relations among concepts to improve KT results. First, LTKT employs a statistical-based approach to construct a learning transfer graph (LTG). Then, LTKT quantifies the impact of an exercise's practice result on the knowledge state of the concept via a direct learning effect component, after which a learning transfer effect component is carefully designed to quantify the impact of the practice result on the knowledge states of neighboring concepts based on the positive and negative learning transfer relations modeled by LTG. We conduct extensive experiments on public real-world datasets, and the experimental results show that LTKT outperforms all state-of-the-art KT methods and has good interpretability. © 2023, The Authors. All rights reserved.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

A Risk-Averse Just-In-Time Scheme for Learning-Based Operation of Microgrids With Coupled Electricity-Hydrogen-Ammonia Under Uncertainties

引用

IEEE Transactions on Sustainable Energy 2025年

作者： Li, Longyan Ning, Chao Pan, Guangsheng Zhang, Leiqi Gu, Wei Zhao, Liang Du, Wenli Shahidehpour, Mohammad Shanghai Jiao Tong University Department of Automation Shanghai200240 China Ministry of Education of China Key Laboratory of System Control and Information Processing Shanghai200240 China School of Electrical Engineering Southeast University Nanjing210096 China Zhejiang Key Laboratory of Distributed Generations and Microgrid Technology State Grid Zhejiang Electric Power Research Institute Hangzhou310014 China Key Laboratory of Smart Manufacturing in Energy Chemical Process Ministry of Education East China University of Science and Technology China Department of Electrical and Computer Engineering Illinois Institute of Technology ChicagoIL60616 United States

This paper proposes a Risk-Averse Just-In-Time (RAJIT) operation scheme for Ammonia-Hydrogen-based Micro-Grids (AHMGs) to boost electricity-hydrogen-ammonia coupling under uncertainties. First, an off-grid AHMG model is developed, featuring a novel multi-mode ammonia synthesis process and a hydrogen-ammonia dual gas turbine with tunable feed-in ratios. Subsequently, a state-behavior mapping strategy linking hydrogen storage levels with the operation modes of ammonia synthesis is established to prevent cost-ineffective shutdowns. The proposed model substantially improves operational flexibility but results in a challenging nonlinear fractional program. Based upon this model, a data-driven RAJIT scheme is developed for the real-time rolling optimization of AHMGs. Unlike conventional one-size-fits-all schemes using one optimization method throughout, the data-driven RAJIT intelligently switches between cost-effective deterministic optimization and risk-averse online-learning distributionally robust optimization depending on actual risk profiles, thus capitalizing on the respective strengths of these two optimization methods. To facilitate the solution of the resulting nonlinear program, we develop an equivalent-reformulation-based solution methodology by leveraging a constraint-tightening technique. Numerical simulations demonstrate that the proposed scheme guarantees safety and yields an overall cost reduction up to 14.6% compared with several state-of-the-art methods. © 2025 IEEE.

关键词： Ammonia

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：