期刊
IEEE TRANSACTIONS ON SERVICES COMPUTING
卷 15, 期 1, 页码 551-566出版社
IEEE COMPUTER SOC
DOI: 10.1109/TSC.2019.2928551
关键词
Deep learning as a service; big data; deep neural networks; accuracy
资金
- US National Science Foundation under CISE SAVI/RCN program [1402266, 1550379]
- US National Science Foundation under CNS program [1421561]
- US National Science Foundation under CRISP program [1541074]
- US National Science Foundation under SaTC program [1564097]
- REU supplement [1545173]
- IBM Faculty Award
- Direct For Computer & Info Scie & Enginr
- Division Of Computer and Network Systems [1550379, 1421561, 1564097] Funding Source: National Science Foundation
- Division Of Computer and Network Systems
- Direct For Computer & Info Scie & Enginr [1541074, 1402266, GRANTS:13872275] Funding Source: National Science Foundation
This paper conducts an empirical comparison and analysis of four representative DL frameworks, highlighting the impact of hyper-parameter configurations on performance and accuracy, as well as the opportunities for improving performance and accuracy through parallel computing library configurations and tuning of hyper-parameters. The study also measures the resource consumption patterns of the DL frameworks and their implications for performance and accuracy. It provides practical guidance for deploying DL as a Service and selecting the right DL frameworks for specific workloads.
Big data powered Deep Learning (DL) and its applications have blossomed in recent years, fueled by three technological trends: a large amount of digitized data openly accessible, a growing number of DL software frameworks in open source and commercial markets, and a selection of affordable parallel computing hardware devices. However, no single DL framework, to date, dominates in terms of performance and accuracy even for baseline classification tasks on standard datasets, making the selection of a DL framework an overwhelming task. This paper takes a holistic approach to conduct empirical comparison and analysis of four representative DL frameworks with three unique contributions. First, given a selection of CPU-GPU configurations, we show that for a specific DL framework, different configurations of its hyper-parameters may have a significant impact on both performance and accuracy of DL applications. Second, to the best of our knowledge, this study is the first to identify the opportunities for improving the training time performance and the accuracy of DL frameworks by configuring parallel computing libraries and tuning individual and multiple hyper-parameters. Third, we also conduct a comparative measurement study on the resource consumption patterns of four DL frameworks and their performance and accuracy implications, including CPU and memory usage, and their correlations to varying settings of hyper-parameters under different configuration combinations of hardware, parallel computing libraries. We argue that this measurement study provides in-depth empirical comparison and analysis of four representative DL frameworks, and offers practical guidance for service providers to deploying and delivering DL as a Service (DLaaS) and for application developers and DLaaS consumers to select the right DL frameworks for the right DL workloads.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据