检索结果-内蒙古大学图书馆

31st International Conference on Computational Linguistics, COLING 2025

作者： Hu, Biao Huang, Zhen Hu, Minghao Yang, Pinglv Qiao, Peng Dou, Yong Wang, Zhilin National Key Laboratory of Parallel and Distributed Computing National University of Defense Technology China Center of Information Research Academy of Military Science China College of Meteorology and Oceanology National University of Defense Technology China

ISBN: (纸本)9798891761964

Prototype network-based methods have made substantial progress in few-shot relation extraction (FSRE) by enhancing relation prototypes with relation descriptions. However, the distribution of relations and instances in distinct representation spaces isolates the constraints of relations on instances, making relation prototypes biased. In this paper, we propose an end-to-end partial order-centered hyperbolic representation learning (PO-HRL) framework, which imposes the constraints of relations on instances by modeling partial order in hyperbolic space, so as to effectively learn the distribution of instance representations. Specifically, we develop the hyperbolic supervised contrastive learning based on Lorentzian cosine similarity to align representations of relations and instances, and model the partial order by constraining instances to reside within the Lorentzian entailment cone of their respective relation. Experiments on three benchmark datasets show that PO-HRL outperforms the strong baselines, especially in 1-shot settings lacking relation descriptions. © 2025 Association for Computational Linguistics.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

PEbfs: Implement High-Performance Breadth-First Search on PEZY-SC3s 24th

PEbfs: Implement High-Performance Breadth-First Search on P...

引用

24th International Conference on Algorithms and Architectures for parallel Processing, ICA3PP 2024

作者： Guo, Weihao Wang, Qinglin Liu, Xiaodong Peng, Muchun Yang, Shun Liang, Yaling Shi, Yongzhen Cao, Ligang Liu, Jie Laboratory of Digitizing Software for Frontier Equipment National University of Defense Technology Changsha410073 China National Key Laboratory of Parallel and Distributed Computing National University of Defense Technology Changsha410073 China Engineering Research Center for National Fundamental Software National University of Defense Technology Changsha410073 China

ISBN: (纸本)9789819615506

The breadth-first search (BFS) algorithm is a fundamental algorithm in graph theory, and it’s parallelization can significantly improve performance. Therefore, there have been numerous efforts to leverage the powerful parallel computing capabilities of hardware like GPGPU to implement high-performance BFS algorithms. However, the energy efficiency is relatively low due to the high power consumption of the platforms on which the algorithm is adapted to. To deal with these challenges, this paper introduces PEbfs that is a high-performance BFS algorithm based on the PEZY-SC3s efficient processor. We integrated three search algorithms, two algorithm optimization strategies, and a directional optimization scheme into PEbfs. Through multiple evaluations of the performance of PEbfs on the public SNAP dataset, the results demonstrate that the average energy efficiency ratio of PEbfs is higher than that of Enterprise and Tigr, the two most advanced implementations on Nvidia’s GPGPU: It achieves 3.08× the average energy efficiency ratio of Enterprise and 4.53× that of Tigr. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

An End-to-End Classification Network Model Based on Hybrid Supervision for Industrial Surface Defect Detection

引用

International Journal of Pattern Recognition and Artificial Intelligence 2025年第5期39卷

作者： Qin, Runbing Chen, Ningjiang Gan, Shukun Guangxi University School of Computer Electronics and Information Nanning530004 China Education Department of Guangxi Zhuang Autonomous Region Key Laboratory of Parallel Distributed and Intelligent Computing Guangxi University Nanning530004 China Guangxi Intelligent Digital Services Research Center of Engineering Technology Nanning530004 China

Industrial part surface defect detection aims to precisely locate defects in images, which is crucial for quality control in manufacturing. The traditional method needs to be designed in advance, but it has shortcomings in terms of generalization ability. Deep neural network-based defect detection faces issues such as low resolution, data scarcity, labeling costs, and high computation. Therefore, an improved solution is necessary to enhance industrial product quality control efficiency and accuracy. A new method is proposed that utilizes end-to-end training of a two-stage neural network based on segmentation with an extended training process. The gradient flow from the classification to the segmentation network is adjusted to prevent unstable features from interfering with learning. To address the problem of image oversampling and undersampling during training, a frequency sampling scheme for negative samples is introduced. Additionally, positive pixels in the region-based segmentation mask are weighted using the distance transform algorithm so that regions with a high probability of defects can be detected without detailed annotation. Experiments are conducted across three distinct defect datasets by applying hybrid supervision encompassing diverse conditions. In the optimal case, the AP rate of the proposed model on the DAGM and KolektorSDD datasets reaches 100%, and the AP rate on the Severstal Steel dataset reaches 98.99%. The experiments’ results show that the detection accuracy has improved greatly, showing how well the proposed method works in the industry. © 2025 World Scientific Publishing Company.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

PERFORMANCE OF parallel SPANNING TREE ALGORITHMS ON LINEAR ARRAYS OF TRANSPUTERS AND UNIX SYSTEMS

引用

parallel computing 1991年第4-5期17卷 527-551页

作者： DAS, SK YANG, CQ Center for Research in Parallel and Distributed Computing Department of Computer Sciences University of North Texas P.O. Box 13886 Denton TX 76203-3886 USA

This paper presents empirical performance of parallel algorithms for connected graphs on the transputer and Unix systems, where processes are configured as a one-dimensional array. We conduct experiments for implementing (i) spanning tree algorithm (SPT-list) with unordered list of edges as input, (ii) SPT algorithm with adjacency matrix, (iii) minimum spanning tree algorithm (MST) with weight matrix as input, and (iv) SPT algorithm derived from MST algorithm by assigning each edge-weight equal to 1. The empirical study is performed with a wide range of random graphs, generated for various (uniformly distributed) edge-densities (d) for a given number (n) of vertices. We plot curves for resulting speedups as functions of n, d, and p. The edge-density d is varied between 0.1 and 0.9;maximum number of vertices (or edges) considered are 300 (or 40 000) and 500 (or 110 000) for transputer and Unix systems, respectively;and the number of processors p varies from 1 through 8 in the transputer system while 1 through 4 in the Unix system. A maximum speedup of 2.98 is obtained on transputers, and that for the Unix system is 3.0. We observe that the speedups of all algorithms vary with increasing number of vertices or edge-density. However, employing more processing units in an algorithm does not necessarily enhance its speedup because of additional communication overhead.

关键词： TRANSPUTERS LINEAR ARRAY PERFORMANCE STUDY parallel GRAPH ALGORITHM SPANNING TREE MINIMUM SPANNING TREE

来源：评论

学校读者我要写书评

暂无评论

On the Complexity of Learning with a Small Number of Nodes

On the Complexity of Learning with a Small Number of Nodes

引用

1992 International Joint Conference on Neural Networks, IJCNN 1992

作者： Parberry, Ian Center for Research in Parallel and Distributed Computing Department of Computer Sciences University of North Texas United States

ISBN: (纸本)0780305590

It is shown that the loading pioblem foi a 6 node neural netwoik with node function set AC^ (that is, the conjunction or disjunction of a subset of the inputs or their complements) is AiV complete. It can be deduced from this observation that the loading problem for a 6 node analog neural network is AfV hard. ©1992 IEEE

关键词：

来源：评论

学校读者我要写书评

暂无评论

Efficient parallel Algorithms for Pattern Recognition

引用

parallel Algorithms and Applications 1994年第1-2期2卷 81-98页

作者： Das, Sajal K Fisher, Paul S Zhang, Hua Center for Research in Parallel and Distributed Computing Department of Computer Science University of North Texas Denton TX 76203-6886 United States

Finitely inductive (F1) sequences are a class of sequences, finite or infinite, which are amenable to a certain mathematical representation which has direct significance to pattern recognition and string matching. Pattern recognition using the Fl technique first transforms known patterns into function table(s) by factoring, and then uses the function table(s) to match an input pattern by following. Factoring is done off line and only once for each pattern. The sequential time for matching two sequences depends only upon the length of the input sequence. In this paper, we propose cost-optimal (i.e., efficient) parallel algorithms for FI sequence processing. These algorithms include parallel factoring and following by bucket packing on an exclusive-read and exclusive-write (EREW), parallel random access machine (PRAM) model. The proposed algorithms also have implications lo parallel integer sorting. © 1994, Taylor & Francis Group, LLC. All rights reserved.

关键词： Finitely inductive sequences Integer sorting parallel algorithms Pattern recognition String matching

来源：评论

学校读者我要写书评

暂无评论

Efficient parallel algorithms on interval graphs 4th

Efficient parallel algorithms on interval graphs

引用

4th International parallel Architectures and Languages Europe Conference, PARLE 1992

作者： Das, Sajal K. Chen, Calvin C. -Y. Center for Research in Parallel and Distributed Computing Department of Computer Science University of North Texas DentonTX76203 United States

ISBN: (纸本)9783540555995

We design efficient parallel algorithms for solving several problems on interval graphs. The problems include finding a BFS-tree and DFS-tree, articulation points and bridges, and minimum coloring. Each of our algorithms requires O(log n) time employing O(n) processors on the EREW PRAM model, where n is the number of vertices. The proposed algorithms for computing articulation points and bridges have better performance in terms of the cost (i.e, processor-time product) than the existing algorithms in [RR90], yet having the same time complexity. Our novel approach to the construction of BFS tree is based on elegantly capturing the structure of a given collection of intervals. This structure reveals important properties of the corresponding interval graph, and is found to be instrumental in solving many other problems on such graphs. Although the time- and processor-complexities of the proposed DFS-tree construction or minimum-coloring is comparable with the best-known ones [KI89], our approach to these problems are new. For example, the approach to the construction of a DFS tree is based on reducing the problem into an all dominating neighbors (ADN) problem;while that to the minimum coloring is by transforming this problem into a linked list ranking problem. © Springer-Verlag Berlin Heidelberg 1992.

关键词： parallel algorithms

来源：评论

学校读者我要写书评

暂无评论

Managing a parallel heap efficiently 3rd

Managing a parallel heap efficiently

引用

3rd International Conference on parallel Architectures and Languages Europe, PARLE 1991

作者： Das, Sajal K. Horng, Wen-Bing Center for Research in Parallel and Distributed Computing Department of Computer Science University of North Texas DentonTX76203-3886 United States

ISBN: (纸本)9783540541516

We design a cost-optimal algorithm for managing a parallel heap on an exclusive-read and exclusive-write (EREW), parallel random access machine (PRAM) model. This is an improvement in space and time over the one recently proposed by Deo and Prasad [4]. Our approach efficiently employs p processors in the range 1 ≤ p ≤ n, where n is the maximum number of items in a parallel heap. It is assumed that a delete-think-insert cycle is repeatedly performed, and each processor requires an arbitrary but the same amount of time (called the think time) for processing an item which in turn generates at most α (a constant) new items. The time required for deleting p items of the highest priority from the parallel heap is O(l), while that for inserting at most αp new items is 0(logn). With or without incorporating the think time, the speedup of our algorithm is proved to be linear, i.e. 0(p). Using a global, working data structure for each level of the heap, it is shown that the additional memory space required for our parallel heap is much less than that for the existing one [4]. Furthermore, the proposed algorithm retains the strict priority ordering of a sequential heap. © Springer-Verlag Berlin Heidelberg 1991.

关键词： Data structures

来源：评论

学校读者我要写书评

暂无评论

Cost-optimal parallel algorithms for finitely inductive sequences 26

Cost-optimal parallel algorithms for finitely inductive sequ...

引用

26th Hawaii International Conference on System Sciences, HICSS 1993

作者： Das, S.K. Fisher, P.S. Zhang, H. Center for Research in Parallel and Distributed Computing Department of Computer Science University of North Texas DentonTX76203-3886 United States

ISBN: (纸本)0818632305

Finitely inductive (FI) sequences are a class of sequences, finite or infinite, which are amenable to a certain mathematical representation which has direct significance to pattern recognition and string matching. Pattern recognition using the FI technique first transforms known patterns into function table(s) by factoring, and then uses the function iable(s) to match an input pattern by following. Factoring is done off line and only once for each pattern. The sequential time for matching two sequences depends only upon the length of the input sequence. In this paper, we propose cost-optimal (i.e., efficient) parallel algorithms for FI sequence processing. These algorithms include parallel factoring and following by bucket packing and by hashing on an exclusive-read and exclusive-write (EREW), parallel random access machine (PRAM) model. © 1993 IEEE.

关键词： Pattern recognition

来源：评论

学校读者我要写书评

暂无评论

Performance of parallel spanning tree algorithms on linear arrays of transputers and Unix systems 5

Performance of parallel spanning tree algorithms on linear a...

引用

5th International parallel Processing Symposium, IPPS 1991

作者： Das, Sajal K. Yang, Cui-Qing Center for Research in Parallel and Distributed Computing Department of Computer Sciences University of North Texas DentonTX76203-3886 United States

ISBN: (纸本)0818691670

This paper presents empirical performance of parallel algorithms for computing a spanning tree (SPT) and a minimum spanning tree (MST) of connected graphs on the Transputer and Unix systems, where processors are configured as a onedimensional array. The parallel MST algorithm uses a weight matrix data structure;and three implementations of the SPT algorithm are presented with unordered edge-list, linked adjacency list and adjacency matrix as data structures. The experiments are conducted with a wide range of random graphs, generated for various edge-densities (d) for a given number (n) of vertices. The edge-density is varied between 0.1 and 0.9, and the maximum number of vertices (or edges) considered are 300 (or 40,000) and 500 (or 110,000) for transputer and Unix systems, respectively. A maximum speed-up of 2.98 is achieved on the transputer network of 8 processors, and that for the Unix system is 3.0 with 4 processes. © 1991 IEEE.

关键词： Transputers

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：