检索结果-内蒙古大学图书馆

6th international conference on parallel, distributed and Grid computing, PDGC 2020

作者： Balabhadrapathruni, Ramya De, Suman S4 PMCIE Oil-Mill SAP Labs India Pvt. Ltd. Karnataka Bangalore India DSC Business Networks AN SAP Labs India Pvt. Ltd. Karnataka Bangalore India

ISBN: (纸本)9781728171326

In recent times, one of the most used scenarios in many industry domains is enhancing the bids or tenders made by suppliers. In this paper, we will be analyzing one such use case for studying the effects of mixed feature selection to optimize the Learning model. The use case is to target and build a predictive clustering model in such a way that the scheduler receives the suggestions based on the most optimal options. There are few feature selection, enhancement, and scaling methodologies which this paper aims to explore with real-time data. Based on the analysis, the most important feature derived would be used to predict the optimal suggestion. The results will then be compared to understand the shortfalls and strong points of this new approach based on the accuracy of prediction. A clustering model will not just help reduce the hours of manual effort put into selecting the right source but will also provide an authentic and optimal option for a scheduler's consideration. © 2020 IEEE.

关键词： Feature Selection

来源：评论

学校读者我要写书评

暂无评论

A distributed Business-Aware Storage Execution Environment Towards Large-Scale Applications 20th

A Distributed Business-Aware Storage Execution Environment T...

引用

20th international conference on Algorithms and Architectures for parallel Processing (ICA3PP)

作者： Jiang, Feng Cheng, Yongyang Dong, Changkun Hui, Zhao Yan, Ruibo China Telecom Cloud Comp Corp Beijing Peoples R China Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9783030602390;9783030602383

The rapid development of distributed storage technology has attracted growing attention from both industry and academia. Although distributed system has been studied for many years, however, it is widely applied to engineering practice since the rise of cloud computing recently. Cloud storage constructs a storage resource pool with massive common storage devices through networks, which could be allocated to authorized users on demand. In this paper, we present a distributed business-aware storage execution environment towards large-scale applications. Initially, we present the overall of our proposed storage execution environment and different data delivery models according to corresponding business needs. Furthermore, we design a globally shared and cross-regional deployed metadata service to minimize the access delay of metadata. Finally, the approach presented in this paper has been validated to be effective through a series of evaluation experiments and actual use cases.

关键词： Cloud computing Object-oriented storage distributed system Large scale applications High-performance architecture

来源：评论

学校读者我要写书评

暂无评论

IEEE Access Special Section: Intelligent Data Sensing, Collection, and Dissemination in Mobile computing

引用

IEEE ACCESS 2020年 8卷 222248-222254页

作者： Liu, Xuxun Liu, Anfeng Tadrous, John He, Ligang Ji, Bo Zheng, Zhongming South China Univ Technol Guangzhou 510641 Peoples R China Cent South Univ Sch Informat Sci & Engn Changsha 410083 Peoples R China Gonzaga Univ Elect & Comp Engn Spokane WA 99258 USA Univ Warwick Coventry CV4 7AL W Midlands England Temple Univ CIS Dept Tokyo 1540004 Japan Temple Univ Ctr Networked Comp CNC Tokyo 1540004 Japan Temple Univ Coll Sci & Technol Dept Comp & Informat Sci Tokyo 1540004 Japan Ericsson Canada Inc Ottawa ON K2K 2V6 Canada

Data are an important parts of building applications and providing services to users. With the development of modern technology, the requirements for data quantity, quality, and processing capabilities have become more stringent. Sufficient high-quality data are the foundation for most mobile applications. It is important to collect enough data to effectively build your application. In addition to ensure the amount of data collected, the quality of the data collected is also important. Otherwise, a large amount of data redundancy will occur, which will cause trouble for data processing. At the same time, large amounts of data also mean the need for data processing capabilities. Traditional processing methods cannot meet the growing demand for data processing. Mobile computing is a promising solution to meet the growing computing needs of various mobile applications. The advancement and application of mobile computing require more intelligent ways to sense and collect data, which constantly poses new challenges to researchers.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Learning Heterogeneous DAG Tasks Scheduling Policies With Efficient Neural Network Evolution

Learning Heterogeneous DAG Tasks Scheduling Policies With Ef...

引用

IEEE international conference on Big Data and Cloud computing (BdCloud)

作者： Weiyu Tan Tian Ling Zhenyu Zeng Ying Tao Yuxia Cheng School of Computer Science Hangzhou Dianzi University Hangzhou China

Neural networks based on NeuroEvolution of Augmenting Topology (NEAT) attracted much attention owing to multiple novel features. It is expected to develop optimal net structure adaptively through evolution and applied to solve specific problems. However, the difficulty of indicating problem status and determining evolution direction along with long execution time makes NEAT hard to show its superiority in many applications. This paper proposed a new DAG scheduling algorithm NeuroSched based on NEAT and parallel Reinforcement Learning. In Markov Decision Processes we use evolved network as a bridge between State of DAG and Action of strategy, and guide network’s evolution direction through Fitness which is calculated by Reward, both of which are under parallel RL environment. Experimental results show that our algorithm is superior to contrasted algorithms in multiple indicators like Makespan and SLR on different scales of DAG, and the parallel process greatly reduces execution time compared with previous evolutionary algorithms.

关键词： Training Scheduling algorithms Network topology Neural networks Reinforcement learning Evolutionary computation Markov processes

来源：评论

学校读者我要写书评

暂无评论

Dynamic Path Based DNN Synergistic Inference Acceleration in Edge computing Environment

Dynamic Path Based DNN Synergistic Inference Acceleration in...

引用

international conference on parallel and distributed Systems (ICPADS)

作者： Meng Zhou Bowen Zhou Huitian Wang Fang Dong Wei Zhao School of Computer Science and Engineering Southeast University Nanjing China Institute for Artficial Intelligence Hefei Comprehensive National Science Center Hefei China School of Computer Science and technology Anhui University of Technology Ma'anshan China

ISBN: (纸本)9781665408790

Deep Neural networks (DNNs) have achieved excellent performance in intelligent applications. Nevertheless, it is elusive for devices with limited resources to support computationally intensive DNNs, while employing the cloud may lead to prohibitive latency. Better solutions are exploiting edge computing and reducing unnecessary computation. Multi-exit DNN based on the early exit mechanism has an impressive effect in the latter, and in edge computing paradigm, model partition on multi-exit chain DNNs is proved to accelerate inference effectively. However, despite reducing computations to some extent, multiple exits may lead to instability of performance due to variable sample quality, performance inferior to the original model especially in the worst case. Furthermore, nowadays DNNs are universally characterized by a directed acyclic graph (DAG), complicating the partition of multi-exit DNN exceedingly. To solve the issues, in this paper, considering online exit prediction and model execution optimization for multi-exit DNN, we propose a Dynamic Path based DNN Synergistic inference acceleration framework (DPDS), where exit designators are designed to avoid iterative entry for exits; to further promote computational synergy in the edge, the multi-exit DNN is dynamically partitioned according to network environment to achieve fine-grained computing offloading. Experimental results show that DPDS can significantly accelerate DNN inference by 1.87× to 6.78×.

关键词： Deep learning Performance evaluation Directed acyclic graph Inference mechanisms Computational modeling Neural networks Predictive models

来源：评论

学校读者我要写书评

暂无评论

AgEBO-Tabular: Joint Neural Architecture and Hyperparameter Search with Autotuned Data-parallel Training for Tabular Data

AgEBO-Tabular: Joint Neural Architecture and Hyperparameter ...

引用

Supercomputing conference

作者： Romain Égelé Prasanna Balaprakash Isabelle Guyon Venkatram Vishwanath Fangfang Xia Rick Stevens Zhengying Liu École polytechnique Palaiseau France Argonne National Laboratory Lemont Illinois USA CNRS/Inria-LISN U. Paris-Saclay France

ISBN: (数字)9781450384421

ISBN: (纸本)9781665483902

Developing high-performing predictive models for large tabular data sets is a challenging task. Neural architecture search (NAS) is an AutoML approach that generates and evaluates multiple neural networks with different architectures concurrently to automatically discover an high performing model. A key issue in NAS, particularly for large data sets, is the large computation time required to evaluate each generated architecture. While data-parallel training has the potential to address this issue, a straightforward approach can result in significant loss of accuracy. To that end, we develop AgEBO-Tabular, which combines Aging Evolution (AE) to search over neural architectures and asynchronous Bayesian optimization (BO) to search over hyperparameters to adapt data-parallel training. We evaluate the efficacy of our approach on two large predictive modeling tabular data sets from the Exascale computing Project-CANcer distributed Learning Environment (ECP-CANDLE).

关键词： Training Exascale computing Distance learning Computational modeling Neural networks distributed databases Computer architecture

来源：评论

学校读者我要写书评

暂无评论

Multi-objective optimization of composing tasks from distributed workflows in cloud computing networks 3rd

Multi-objective optimization of composing tasks from distrib...

引用

3rd international conference on Computational Intelligence and Informatics, ICCII 2018

作者： Mohan, V. Murali Satyanarayana, K.V.V. Department of Computer Science and Engineering Koneru Lakshmaiah Education Foundation GunturAP522502 India

ISBN: (纸本)9789811514791

This manuscript proposed and explored a novel strategy for optimizing the composition of tasks from distributed workflows to achieve parallel execution, optimality toward resource utilization. The critical objective of the proposal is to optimize the task sequences from the workflows initiated to execute parallel in distributed RDF environment, which is unique regard to the earlier contributions related to parallel query planning and execution strategies found in contemporary literature. All of these existing models are aimed to notify the tasks from the given workflow, which are less significant to optimize the parallel execution of the multiple tasks located in distributed workflows. In order to this, the multi-objective optimization of composing task (MOCT) sequences from distributed workflows is proposed. The MOCT optimizes the execution of tasks from multiple workflows initiated in parallel. A novel scale called "task sequence consistency score" uses the order of other metric task sequence coverages, dependency scope, and lifespan as input. The experiments are conducted on the proposed model, and other benchmark models are found in contemporary literature. The results are obtained from the experimental study evincing that the MOCT is significant and robust to optimize the task sequences in order to execute distribute workflows in parallel. The comparative analysis of the results is obtained from MOCT, and other contemporary models are performed using ANOVA standards like t-test and Wilcoxon signed-rank test. © Springer Nature Singapore Pte Ltd. 2020.

关键词： Multiobjective optimization

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

NUMA-aware FFT-based Convolution on ARMv8 Many-core CPUs

引用

IEEE international conference on Big Data and Cloud computing (BdCloud)

作者： Xiandong Huang Qinglin Wang Shuyu Lu Ruochen Hao Songzhu Mei Jie Liu Science and Technology on Parallel and Distributed Processing Laboratory National University of Defense Technology Changsha China School of Computer Science National University of Defense Technology Changsha China University of Pittsburgh Pittsburgh USA

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 many-core 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.

关键词： Deep learning Recurrent neural networks Convolution Computer architecture Complexity theory Convolutional neural networks Kernel

来源：评论

学校读者我要写书评

暂无评论

Energy-Efficient CNNs Accelerator Implementation on FPGA with Optimized Storage and Dataflow

Energy-Efficient CNNs Accelerator Implementation on FPGA wit...

引用

IEEE international conference on Big Data and Cloud computing (BdCloud)

作者： Yonghua Zhang Hongxu Jiang Xiaobin Li Rui Miao Jinyan Nie Yu Du Beijing Key Laboratory of Digital Media of Beihang University Hangzhou Innovation Institute of Beihang University Beijing China Beihang University Beijing China Beijing Union University Beijing China

Convolutional neural networks (CNNs) has been widely used in computer vision and speech processing, and has achieved great success. However, the deployment of large-scale CNN model is limited by computing and memory in the smart embedded system. Through the current high parallel computing paradigm of CNNs accelerator, the computing requirements can be effectively met to achieve high throughput. However, because the communication cost may be higher than the computing cost for small smart platform, the energy consumption is still very high. In order to solve this problem, a new CNNs accelerator based on storage and data flow is proposed, which realizes energy-saving CNNs inference acceleration by minimizing data access and maximizing data reuse. This paper implements the accelerator on the Zynq UltraScale+ MPSoC ZCU102 evaluation board, and evaluates the throughput and energy efficiency of the accelerator for typical vgg16 and tiny Yolo benchmark networks. Compared with other accelerators, the our accelerator improves the system energy efficiency by 6.3×, the system throughput by 41×, and the throughput of a single DSP by 7.63×.

关键词： Energy consumption Costs Embedded systems parallel processing Throughput Energy efficiency System-on-chip

来源：评论

学校读者我要写书评

暂无评论

A Survey on Privacy Approaches for Social networks

A Survey on Privacy Approaches for Social Networks

引用

IEEE international conference on Big Data and Cloud computing (BdCloud)

作者： Regina Marin Luciano Gallegos Instituto Federal Goiano (IFGoiano) Ceres Brazil Data Science and Artificial Intelligence Laboratory University of Fortaleza (UNIFOR) Fortaleza Brazil

The popular use of Social networks (SNs) for online messaging and information exchange provided available online massive amount of personal data. In general, social interactions require information sharing between involved parties, which implies knowing with whom, and to which extent the information is shared. In this paper, we provide a comprehensive overview of SNs state-of-the-art by exploring users’ privacy in SNs implementations from centralized to fully decentralized architecture. To clarify whether existing techniques can help to prevent attacks or not, we survey and compare how privacy policies are specified and operationalized into the SN design. The result of our in-depth analysis suggests that currently privacy enforcement mechanisms are not satisfactory in SNs because they produce unbalanced control of the shared information.

关键词： Privacy Data privacy Social networking (online) Information sharing Information exchange

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：