检索结果-内蒙古大学图书馆

GroPipe: A Grouped Pipeline Hybrid Parallel Method for Accelerating DCNNs Training

学校读者我要写书评

暂无评论

IEEE Transactions on computers 2025年第7期74卷 2487-2500页

作者： Liu, Bin Ma, Yongyao Hu, Zijian Ji, Zeyu He, Zhenli Li, Keqin Northwest A&F University College of Information Engineering Shaanxi Yangling712100 China Shaanxi Engineering Research Center of Agricultural Information Intelligent Perception and Analysis Shaanxi Key Laboratory of Agricultural Information Perception and Intelligent Service Key Laboratory of Agricultural Internet of Things Ministry of Agriculture and Rural Affairs China Yunnan University Yunnan Key Laboratory of Software Engineering Kunming650091 China Yunnan University School of Software Kunming650091 China State University of New York Department of Computer Science New PaltzNY12561 United States

Training large Deep Convolutional Neural Networks (DCNNs) with increasingly large datasets to improve model accuracy has become extremely time-consuming. Distributed training methods, such as data parallelism (DP) and pipeline model parallelism (PMP), offer potential solutions but face challenges like load imbalance and significant communication overhead. This paper introduces GroPipe, a novel architecture that synergistically integrates PMP and DP, markedly improving training speeds. GroPipe employs an automatic model partitioning algorithm based on a performance projection technique, ensuring load balance and facilitating quantitative performance evaluation in PMP. Additionally, it adopts a group-based delayed asynchronous communication strategy to efficiently reduce communication overhead in DP. Using the ResNet and VGG models with the ImageNet dataset, extensive experiments are performed on an 8-GPU server and demonstrate GroPipe’s effectiveness. GroPipe achieves substantial improvements in time to accuracy, showing an average improvement of 42.2% and 14.0% on the ResNet series, and 79.2% and 43.9% on the VGG series, without compromising Top-1 accuracy. © 1968-2012 IEEE.

关键词： Deep convolutional neural networks Delayed asynchronous communication Hybrid parallelism Pipelined model parallelism

Probery: A probability-based incomplete query optimization for big data

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Song, Jie Zhang, Yichuan Bao, Yubin Yu, Ge Software College Northeastern University School of Computer Science and Engineering Northeastern University

Nowadays, query optimization has been highly concerned in big data management, especially in NoSQL databases. Approximate queries boost query performance by loss of accuracy, for example, sampling approaches trade off query completeness for efficiency. Different from them, we propose an uncertainty of query completeness, called Probability of query Completeness (PC for short). PC refers to the possibility that query results contain all satisfied records. For example PC=0.95, it guarantees that there are no more than 5 incomplete queries among 100 ones, but not guarantees how incomplete they are. We trade off PC for query performance, and experiments show that a small loss of PC doubles query performance. The proposed Probery (PROBability-based data quERY) adopts the uncertainty of query completeness to accelerate OLTP queries. This paper illustrates the data and probability models, the probability based data placement and query processing, and the Apache Drill-based implementation of Probery. In experiments, we first prove that the percentage of complete queries is larger than the given PC confidence for various cases, namely that the PC guarantee is validate. Then Probery is compared with Drill, Impala and Hive in terms of query performance. The results indicate that Drill-based Probery performs as fast as Drill with complete query, while averagely 1.8x, 1.3x and 1.6x faster than Drill, Impala and Hive with possible complete query, respectively. Copyright © 2019, The Authors. All rights reserved.

关键词： Probability

IMSpec: An Extensible Approach to Exploring the Incorrect Usage of APIs

学校读者我要写书评

暂无评论

IMSpec: An Extensible Approach to Exploring the Incorrect Us...

Theoretical Aspects of software engineering (TASE)

作者： Zuxing Gu Min Zhou Jiecheng Wu Yu Jiang Jiaxiang Liu Ming Gu School of Software Tsinghua University China College of Computer Science & Software Engineering Shenzhen University China

Application Programming Interfaces (APIs) usually have usage constraints, such as call conditions or call orders. Incorrect usage of these constraints, called API misuse, will result in system crashes, bugs, and even security problems. It is crucial to detect such misuses early in the development process. Though many approaches have been proposed over the last years, recent studies show that API misuses are still prevalent, especially the ones specific to individual projects. In this paper, we strive to improve current API-misuse detection capability for large-scale C programs. First, We propose IMSpec, a lightweight domain-specific language enabling developers to specify API usage constraints in three different aspects (i.e., parameter validation, error handling, and causal calling), which are the majority of API-misuse bugs. Then, we have tailored a constraint guided static analysis engine to automatically parse IMSpec rules and detect API-misuse bugs with rich semantics. We evaluate our approach on widely used benchmarks and real-world projects. The results show that our easily extensible approach performs better than state-of-the-art tools. We also discover 19 previously unknown bugs in real-world open-source projects, all of which have been confirmed by the corresponding developers.

关键词： computer bugs Static analysis Syntactics Semantics software Tools Engines

Addressing time bias in bipartite graph ranking for important node identification

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Liao, Hao Wu, Jiao Zhou, Mingyang Vidmer, Alexandre College of Computer Science and Software Engineering Shenzhen University Shenzhen518060 China

The goal of the ranking problem in networks is to rank nodes from best to worst, according to a chosen criterion. In this work, we focus on ranking the nodes according to their quality. The problem of ranking the nodes in bipartite networks is valuable for many real-world applications. For instance, high quality products can be promoted on an online shop or highly reputed restaurants attract more people on venues reviews platforms. However, many classical ranking algorithms share a common drawback: they tend to rank older movies higher than newer movies, though some newer movies may have a high quality. This time bias originates from the fact that older nodes in a network tend to have more connections than newer ones. In the study, we develop a ranking method using a rebalance approach to diminish the time bias of the rankings in bipartite graphs. As a result, it is indicated that the rankings produced after applying the rebalancing of the scores not only reduce time bias, but also reflect the quality of the benchmark items more accurately. The ranking performance improvements range from 20% to 80%, the experiments were conducted on three real datasets with ground truth benchmark. Copyright © 2019, The Authors. All rights reserved.

关键词： Graph theory

Adaptive layer-wise personalized federated learning via dual delay update in future communication networks

学校读者我要写书评

暂无评论

computer Communications 2025年 239卷

作者： Xu, Haowen Mao, Yingchi Chen, Si Rong, Yi Muazu, Tasiu He, Xiaoming College of Computer Science And Software Engineering Hohai University No. 8 Focheng West Road Jiangning District Nanjing211100 China Key Laboratory of Water Big Data Technology of Ministry of Water Resources Hohai University Nanjing210098 China China Huaneng Group Co. Ltd Beijing100000 China College of Internet of Things Nanjing University of Posts and Telecommunications Nanjing210003 China

The future communication networks refers to large-scale mass-connected networks consisting of billions of cloud, edge, and end devices, which are expected to support the ever-growing communication demands. In the future communication networks, billions of end devices generate massive amount of data that needs to be processed and analyzed (e.g., model training). Artificial Intelligence of Things (AIoT) is a groundbreaking technology that leverages artificial intelligence models to process and analyze data generated by a large number of internet of things devices. As an emerging AIoT method, personalized Federated Learning (pFL) has emerged prominently in distributed model training using massive data from the future communication networks. However, it is challenging to accomplish high-performance and communication-efficient model training by existing pFL methods in the future communication networks, due to the following limitations. a) Dynamic role differences in each layer of a multi-layer model are neglected, leading to poor accuracy in customized models deployed on end devices. b) Owing to numerous end devices in the future communication networks, the communication frequency between a cloud server and end devices is extremely high in each communication round, resulting in expensive communication cost. To solve these two limitations, this paper presents a novel pFL framework for distributed model training in the future communication networks, called Adaptive Layer-wise personalized Federated Learning via Dual Delay Update (ALpFLDDU). First, in end devices, a layer-wise aggregation scheme based on an adaptive weight calculation mechanism is designed to capture the dynamic role differences of model layers. Second, in each communication round, we develop a dual delay update strategy to reduce communication frequency between a cloud server and end devices while ensuring model performance. Simulation experiments on text and image classification datasets are conducted. Th

关键词： Internet of things

FedMint: Intelligent Bilateral Client Selection in Federated Learning with Newcomer IoT Devices

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Wehbi, Osama Arisdakessian, Sarhad Wahab, Omar Abdel Otrok, Hadi Otoum, Safa Mourad, Azzam Guizani, Mohsen Cyber Security Systems and Applied AI Research Center Department of CSM Lebanese American University Lebanon Department of Computer and Software Engineering Polytechnique Montréal MontrealQC Canada Khalifa University Abu Dhabi United Arab Emirates College of Technological Innovation Zayed University Dubai United Arab Emirates Division of Science New York University Abu Dhabi United Arab Emirates Mohammad Bin Zayed University of Artificial Intelligence Abu Dhabi United Arab Emirates

Federated Learning (FL) is a novel distributed privacy-preserving learning paradigm, which enables the collaboration among several participants (e.g., Internet of Things devices) for the training of machine learning models. However, selecting the participants that would contribute to this collaborative training is highly challenging. Adopting a random selection strategy would entail substantial problems due to the heterogeneity in terms of data quality, and computational and communication resources across the participants. Although several approaches have been proposed in the literature to overcome the problem of random selection, most of these approaches follow a unilateral selection strategy. In fact, they base their selection strategy on only the federated server’s side, while overlooking the interests of the client devices in the process. To overcome this problem, we present in this paper FedMint, an intelligent client selection approach for federated learning on IoT devices using game theory and bootstrapping mechanism. Our solution involves the design of: (1) preference functions for the client IoT devices and federated servers to allow them to rank each other according to several factors such as accuracy and price, (2) intelligent matching algorithms that take into account the preferences of both parties in their design, and (3) bootstrapping technique that capitalizes on the collaboration of multiple federated servers in order to assign initial accuracy value for the newly connected IoT devices. Based on our simulation findings, our strategy surpasses the VanillaFL selection approach in terms of maximizing both the revenues of the client devices and accuracy of the global federated learning model. © 2022, CC BY-NC-SA.

关键词： Game theory

Medical question similarity calculation based on weighted domain dictionary 18

学校读者我要写书评

暂无评论

Medical question similarity calculation based on weighted do...

2018 International Conference on Big Data and Computing, ICBDC 2018

作者： Yan, Guokai Li, Jianqiang College of Computer Science and Software Engineering Shenzhen University China

ISBN: (纸本)9781450364263

In the calculation of question similarity in the mobile medical question answering system, not all the importance of the words are the same. To solve the problem, we proposed a improved similarity calculation method based on the vector space model and combining the weighted domain dictionary for mobile health questions in question and answer system. Adding weight to field word in target question to bring a higher scores of similarity calculation, reduce the impact of non-field word in similarity calculation. © 2018 Association for Computing Machinery.

关键词： Vector spaces

HSNP-Miner: High Utility Self-Adaptive Nonoverlapping Pattern Mining

学校读者我要写书评

暂无评论

HSNP-Miner: High Utility Self-Adaptive Nonoverlapping Patter...

IEEE International Conference on Big Knowledge (ICBK)

作者： Md Motaher Hossain Youxi Wu Philippe Fournier-Viger Zhao Li Lei Guo Yan Li School of Artificial Intelligence Hebei University of Technology Tianjin China College of Computer Science and Software Engineering Shenzhen University Shenzhen China Alibaba Group Hangzhou Zhejiang China State Key Laboratory of Reliability and Intelligence of Electrical Equipment Hebei University of Technology Tianjing China School of Economics and Management Hebei University of Technology Tianjing China

ISBN: (纸本)9781665438599

Sequential pattern mining (SPM) under the nonoverlapping condition (or nonoverlapping SPM) is a type of data mining used to extract frequent gapped subsequences (known as patterns) from sequences, which is more valuable and versatile than other related methods. In nonoverlapping SPM, two occurrences cannot reuse the same sequence letter in the exact location as the occurrences. This method evaluates the frequency of the patterns in the sequence, and ignores the impact of external utility (item price or profit). Therefore, some low-frequency and essential patterns are overlooked. To address this issue, this paper introduces High Utility Self-adaptive Nonoverlapping Pattern (HSNP) mining and proposes HSNP-Miner, which includes two steps: support calculation and candi-date pattern generation. To calculate the support, we propose the NoSup algorithm, which can effectively calculate support while avoiding the creation of redundant nodes. An advanced upper bound method is employed to generate the candidate patterns more efficiently. Compared to other competitive methods, the experimental results demonstrate the efficiency of the proposed algorithm and the uniqueness of nonoverlapping sequence pat-tarns.

关键词： Upper bound Conferences DNA Data mining Viruses (medical)

Energy efficiency and hover time optimization in UAV-based hetnets

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Muntaha, Sidra Tul Hassan, Syed Ali Jung, Haejoon Hossain, M. Shamim School of Electrical Engineering and Computer Science National University of Sciences and Technology Islamabad44000 Pakistan Department of Information and Telecommunication Engineering Incheon National University Yeonsugu Incheon Korea Republic of Department of Software Engineering College of Computer and Information Sciences King Saud University Riyadh11543 Saudi Arabia

In this paper, we investigate the downlink performance of a three-tier heterogeneous network (HetNet). The objective is to enhance the edge capacity of a macro cell by deploying unmanned aerial vehicles (UAVs) as flying base stations and small cells (SCs) for improving the capacity of indoor users in scenarios such as temporary hotspot regions or during disaster situations where the terrestrial network is either insufficient or out of service. UAVs are energy-constrained devices with a limited flight time, therefore, we formulate a two layer optimization scheme, where we first optimize the power consumption of each tier for enhancing the system energy efficiency (EE) under a minimum quality-of-service (QoS) requirement, which is followed by optimizing the average hover time of UAVs. We obtain the solution to these nonlinear constrained optimization problems by first utilizing the Lagrange multipliers method and then implementing a sub-gradient approach for obtaining convergence. The results show that through optimal power allocation, the system EE improves significantly in comparison to when maximum power is allocated to users (ground cellular users or connected vehicles). The hover time optimization results in increased flight time of UAVs thus providing service for longer durations. Copyright © 2020, The Authors. All rights reserved.

关键词： Energy efficiency