检索结果-内蒙古大学图书馆

IEEE Global Communications Conference (GLOBECOM) on Advanced Technology for 5G Plus

作者： Xu, Jiasheng Fu, Luoyi Wang, Xinbing Shanghai Jiao Tong Univ Dept Elect Engn Shanghai Peoples R China Shanghai Jiao Tong Univ Dept Comp Sci Shanghai Peoples R China

ISBN: (纸本)9781728182988

distributed computing is known for its high efficiency of processing large amounts of data in parallel, at the expense of communication load between different servers. Coding was introduced to minimize the communication load by exploiting the repetitive computing, thus drawing great attention within the academia. Most existing works assume that all servers are identical in computational capability, which is inconsistent with practical scenarios. In this paper, we investigate a distributed computing system that consists of two types of servers, i.e., fast servers and slow servers. Due to the heterogeneous computational capabilities within the system, the overall computation time will be delayed by the slow servers, which is called the straggling effect. To this end, we develop a novel framework of coding-based distributed computing to alleviate the straggling effect. Specifically, for a given number of fast servers and slow servers with their corresponding computational capabilities, we aim to minimize the overall computation time by assigning different amounts of workloads to different servers. Further, we derive the information-theoretic lower hound of the communication load of the system, which is shown to be within a constant multiplicative gap to the achievable communication load by our scheme.

关键词： distributed computing coded computation heterogeneous servers straggling effect

来源：评论

学校读者我要写书评

暂无评论

Decision Power of Weak Asynchronous Models of distributed computing 21

Decision Power of Weak Asynchronous Models of Distributed Co...

引用

40th ACM SIGACT-SIGOPS Symposium on Principles of distributed computing (PODC)

作者： Czerner, Philipp Guttenberg, Roland Helfrich, Martin Esparza, Javier Tech Univ Munich Munich Germany

ISBN: (纸本)9781450385480

Esparza and Reiter have recently conducted a systematic comparative study of models of distributed computing consisting of a network of identical finite-state automata that cooperate to decide if the underlying graph of the network satisfies a given property. The study classifies models according to four criteria, and shows that twenty-four initially possible combinations collapse into seven equivalence classes with respect to their decision power, i.e. the properties that the automata of each class can decide. However, Esparza and Reiter only show (proper) inclusions between the classes, and so do not characterise their decision power. In this paper we do so for labelling properties, i.e. properties that depend only on the labels of the nodes, but not on the structure of the graph. In particular, majority (whether more nodes carry label a than b) is a labelling property. Our results show that only one of the seven equivalence classes identified by Esparza and Reiter can decide majority for arbitrary networks. We then study the expressive power of the classes on bounded-degree networks, and show that three classes can. In particular, we present an algorithm for majority that works for all bounded-degree networks under adversarial schedulers, i.e. even if the scheduler must only satisfy that every node makes a move infinitely often, and prove that no such algorithm can work for arbitrary networks.

关键词： distributed computing graph automata weak asynchronous models labelling properties communication graphs decision power

来源：评论

学校读者我要写书评

暂无评论

Coded computing for Master-Aided distributed computing Systems

Coded Computing for Master-Aided Distributed Computing Syste...

引用

IEEE Information Theory Workshop (ITW)

作者： Chen, Haoning Wu, Youlong ShanghaiTech Univ Sch Informat Sci & Technol Shanghai Peoples R China

ISBN: (纸本)9781728159621

We consider a MapReduce-type task running in a distributed computing model which consists of K edge computing nodes distributed across the edge of the network and a Master node that assists the edge nodes to compute output functions. The Master node and the edge nodes, both equipped with some storage memories and computing capabilities, are connected through a multicast network. We define the communication time spent during the transmission for the sequential implementation (all nodes send symbols sequentially) and parallel implementation (the Master node can send symbols during the edge nodes' transmission), respectively. We propose a mixed coded distributed computing scheme that divides the system into two subsystems where the coded distributed computing (CDC) strategy proposed by Songze Li et al. is applied into the first subsystem and a novel master-aided CDC strategy is applied into the second subsystem. We prove that this scheme is optimal, i.e., achieves the minimum communication time for both the sequential and parallel implementation, and establish an optimal information-theoretic tradeoff between the overall communication time, computation load, and the Master node's storage capacity. It demonstrates that incorporating a Master node with storage and computing capabilities can further reduce the communication time. For the sequential implementation, we deduce the approximately optimal file allocation between the two subsystems, which shows that the Master node should map as many files as possible in order to achieve smaller communication time. For the parallel implementation, if the Master node's storage and computing capabilities are sufficiently large (not necessary to store and map all files), then the proposed scheme requires at most 1/2 of the minimum communication time of system without the help of the Master node.

关键词： distributed computing MapReduce communication computation

来源：评论

学校读者我要写书评

暂无评论

Chinese keyword extraction model with distributed computing

引用

COMPUTERS & ELECTRICAL ENGINEERING 2022年 97卷 107639-107639页

作者： Ding, Tiantian Yang, Wenzhong Wei, Fuyuan Ding, Chao Kang, Peng Bu, Wenxiu Xinjiang Univ Coll Informat Sci & Engn Urumqi 830046 Xinjiang Peoples R China Xinjiang Univ Coll Informat Sci & Engn Xinjiang Lab Multilanguage Informat Technol Urumqi 830046 Xinjiang Peoples R China

The content and structure of the news text are relatively complex and cannot effectively capture the core content. Existing supervision models cannot achieve good results in areas with less annotated data. To this end, we propose a new Chinese keyword extraction model with a distributed computing method. Precisely, we first fused the Bidirectional Encoder Representation from Transformers (BERT) and Conditional Random Fields (CRF) so that each word learns its relationship with the context while reducing errors;secondly, the adversarial training encourages the model to retain a small amount of annotations Sample knowledge to help extract keywords from unannotated samples;and because the model contains a large number of time-consuming components, it creatively uses distributed computing to save overall computing time. The results show that our model can steadily improve the performance of keyword phrase extraction in areas with insufficient labeled samples.

关键词： Keyword Extraction BERT Adversarial training CRF distributed computing

来源：评论

学校读者我要写书评

暂无评论

A solar forecasting framework based on federated learning and distributed computing

引用

BUILDING AND ENVIRONMENT 2022年第0期225卷

作者： Wen, Haoran Du, Yang Lim, Eng Gee Wen, Huiqing Yan, Ke Li, Xingshuo Jiang, Lin Xian Jiaotong Liverpool Univ Sch Adv Technol Suzhou Peoples R China James Cook Univ Coll Sci & Engn Douglas Australia Natl Univ Singapore Dept Built Environm Singapore Singapore Nanjing Normal Univ Sch Elect & Automation Engn Nanjing Peoples R China Univ Liverpool Dept Elect & Elect Liverpool England

Solar forecasting is a crucial and cost-effective tool for better utilization of solar energy for smart environment design. Artificial intelligence (AI) technologies, such as machine learning (ML) and deep learning (DL), have gained great popularity and been widely applied in solar forecasting in recent years. However, conventional AI -based forecasting methods suffer from high variability and stochasticity of solar irradiation, making unreliable predictions due to heterogeneous solar resources. Moreover, the training process of DL models is less flexible and requires immense data. Even for a well-trained model, it can still yield deteriorated performances on other datasets of varying data distributions. To tackle the deficiencies of AI forecasting models, we present a flexible distributed solar forecasting framework based on a novel spatial and temporal attention-based neural network (STANN) in conjunction with federated learning (FL) technique, considering multi-horizon forecasting scenario from 5-30 min. The STANN model consists of a feature extractor and a forecaster, which can be respectively trained on various local datasets for better localization, and updated to further improve forecasting accuracy through global parameter aggregation under the proposed framework without data gathering. We evaluate effectiveness of the proposed method by conducting extensive experiments on real-world datasets and compare it to other popular forecasting models. The results demonstrate that our approach outperforms the other benchmarks with higher forecasting accuracy for all forecast horizons and better generalization on various datasets, achieving the highest forecast skill of 28.83% at 30 min horizon and an improvement of 11.2% compared with the centralized, localized, and conventional FL training methods.

关键词： Solar forecasting Deep learning Federated learning distributed computing Sustainable environment

来源：评论

学校读者我要写书评

暂无评论

Robust and Efficient Algorithms for Federated Learning and distributed computing

Robust and Efficient Algorithms for Federated Learning and D...

引用

作者： Reisizadehmobarakeh, Amirhossein University of California Santa Barbara

学位级别：Ph.D., Doctor of Philosophy

Training a large-scale model over a massive data set is an extremely computation and storage intensive task, e.g. training ResNet with hundreds of millions of parameters over the data set ImageNet with millions of images. As a result, there has been significant interest in developing distributed learning strategies that speed up the training of learning models. Due to the growing computational power of the ecosystem of billions of mobile and computing devices, many future distributed learning systems operate based on storing data locally and pushing computation to the network edge. Unlike traditional centralized machine learning environments, however, machine learning at the edge is characterized by significant challenges including (1) scalability due to severe constraints on communication bandwidth and other resources including storage and energy, (2) robustness to stragglers, and edge failures due to slow edge nodes, (3) models generalizing to non-i.i.d. and heterogeneous data. In this thesis, we focus on two important distributed learning frameworks: Federated Learning and distributed computing, with a shared goal in mind: how to provably address the critical challenges in such paradigms using novel techniques from distributed optimization, statistical learning theory, probability theory, and communication and coding theory to advance the state-of-the-art in efficiency, resiliency, and scalability. In the first part of the thesis, we devise three methods to mitigate communication cost, straggler resiliency and robustness to heterogeneous data in federated learning paradigms. Our main ideas are to employ model compression, adaptive device participation and distributionally robust minimax optimization, respectively for such challenges. We characterize provable improvements for the proposed algorithms in terms of convergence speed, expected runtime, and generalization gaps. Moving on to the second part, we consider important instances of distributed computing framew

关键词： Coded computing distributed computing distributed optimization Federated learning Machine learning environments

来源：评论

学校读者我要写书评

暂无评论

A Semantic Data-Based distributed computing Framework to Accelerate Digital Twin Services for Large-Scale Disasters

引用

SENSORS 2022年第18期22卷 6749-6749页

作者： Kwon, Jin-Woo Yun, Seong-Jin Kim, Won-Tae Korea Univ Technol & Educ Dept Comp Sci & Engn Future Convergence Engn Major Cheonan 31253 South Korea

As natural disasters become extensive, due to various environmental problems, such as the global warming, it is difficult for the disaster management systems to rapidly provide disaster prediction services, due to complex natural phenomena. Digital twins can effectively provide the services using high-fidelity disaster models and real-time observational data with distributed computing schemes. However, the previous schemes take little account of the correlations between environmental data of disasters, such as landscapes and weather. This causes inaccurate computing load predictions resulting in unbalanced load partitioning, which increases the prediction service times of the disaster management agencies. In this paper, we propose a novel distributed computing framework to accelerate the prediction services through semantic analyses of correlations between the environmental data. The framework combines the data into disaster semantic data to represent the initial disaster states, such as the sizes of wildfire burn scars and fuel models. With the semantic data, the framework predicts computing loads using the convolutional neural network-based algorithm, partitions the simulation model into balanced sub-models, and allocates the sub-models into distributed computing nodes. As a result, the proposal shows up to 38.5% of the prediction time decreases, compared to the previous schemes.

关键词： balanced partitioning computing load prediction digital twin distributed computing large-scale disasters load balancing machine learning semantic data

来源：评论

学校读者我要写书评

暂无评论

A parking space allocation algorithm based on distributed computing

引用

INTERNATIONAL JOURNAL OF SENSOR NETWORKS 2020年第4期33卷 250-258页

作者： Chen, Guanlin Pang, Huajian Xu, Huang Yang, Wujian Chen, Yong Zhejiang Univ City Coll Sch Comp & Comp Sci Hangzhou 310015 Peoples R China Changzhou Univ Sch Informat Sci & Engn Changzhou 213164 Jiangsu Peoples R China Zhejiang Univ City Coll Sch Business Hangzhou 310015 Peoples R China

In order to make it easier for drivers to find a parking slot, optimise the resources of urban parking slots, and alleviate the problem of parking slot shortage, a distributed parking allocation algorithm was proposed. The algorithm collects the parking requests of user, this parking requests including the current position coordinate information of the users and destination coordinate information, the algorithm allocates parking spaces to users by analysing the available state of parking spaces, then return the parking route planning to the client. Compared with the traditional algorithm, the distributed parking algorithm has a higher ability to withstand pressure and global search capability, and it can ensure the real-time and validity of the parking information, so it can reduce the problem of the parking space shortage and unavailable parking space. The simulation results show that this algorithm can find the solution set more quickly and accurately under the circumstance of high demand. It also has application value and practicality.

关键词： city traffic parking allocation distributed computing matching algorithm smart city genetic algorithm parking slot urban management parking requests greedy algorithm

来源：评论

学校读者我要写书评

暂无评论

Parallel multi-view concept clustering in distributed computing

引用

NEURAL computing & APPLICATIONS 2020年第10期32卷 5621-5631页

作者： Wang, Hao Yang, Yan Zhang, Xiaobo Peng, Bo Southwest Jiaotong Univ Sch Informat Sci & Technol Chengdu 65576 Peoples R China

Multi-view clustering (MvC) is an emerging task in data mining. It aims at partitioning the data sampled from multiple views. Although a great deal of research has been done, this task remains to be very challenging. We found an important problem in performing the MvC task. MvC needs large amounts of computation. To address this problem, we propose a parallel MvC method in a distributed computing environment. The proposed method builds upon concept factorization with local manifold learning, denoted by parallel multi-view concept clustering (PMCC). Concept factorization learns a compressed representation for the data. Local manifold learning preserves the locally intrinsic geometrical structure in the data. The weight of each view is learned automatically and a cooperative normalized approach is proposed to better guide the learning of a consensus representation for all views. For the proposed PMCC architecture, the calculation of each part is independent. It is clear that our PMCC can be performed in a distributed computing environment. Experimental results using real-world datasets demonstrate the effectiveness of the proposed method.

关键词： Multi-view clustering Concept factorization Manifold learning distributed computing

来源：评论

学校读者我要写书评

暂无评论

Blockchain and federated learning-based distributed computing defence framework for sustainable society

引用

SUSTAINABLE CITIES AND SOCIETY 2020年 59卷 102220-102220页

作者： Sharma, Pradip Kumar Park, Jong Hyuk Cho, Kyungeun Dongguk Univ Dept Multimedia Engn Seoul 04620 South Korea Seoul Natl Univ Sci & Technol Dept Dept Comp Sci & Engn Seoul 01811 South Korea

Ensuring social security through the defense organization determines the creation of links between the army and society. Realizing the benefits of the Internet of Battle Things in the defense system can perfectly monetize intelligence and strengthen the armed forces. It establishes a network for strong connectivity in the army with good coordination between complex processes to effectively edge out the enemies. However, this new technology poses organizational and national security challenges that present both opportunities and obstacles. The current framework of the defense IoT network for sustainable society is not adequate enough to make actionable situational awareness decisions in order to infer the state of the battlefield while preserving the privacy of sensitive data. In this paper, we propose a distributed computing defence framework for sustainable society using the features of blockchain technology and federated learning. The proposed model presents an algorithm to meet the challenges of limited training data in order to obtain high accuracy and avoid a reason specific model. To evaluate the effectiveness of the proposed model, we prepare the dataset and investigate the performance of our framework in various scenarios. The result outcomes are promising in terms of accuracy and loss compared to baseline approach.

关键词： distributed computing Internet of battle things Sustainable society Blockchain Federated learning

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：