权益分类	功能权益	普通用户	{{item.name}}会员
{{category.name}}	{{benefitItem.name}}

CSR: Medium: Collaborative Research: Scale-Out Near-Data Acceleration of Machine Learning

CSR：媒介：协作研究：机器学习的横向扩展近数据加速

基本信息

批准号：
1833373
负责人：
Hadi Esmaeilzadeh
金额：
$ 50万
依托单位：
University of California-San Diego
依托单位国家：
美国
项目类别：
Continuing Grant
财政年份：
2018
资助国家：
美国
起止时间：
2018-02-07 至 2023-09-30
项目状态：
已结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=1833373&HistoricalAwards=false
关键词：
CSR Medium Collaborative Research Scale

项目摘要

A growing number of commercial and enterprise systems increasingly rely on machine learning algorithms. This shift is, on the one hand, due to the breakthroughs in machine learning algorithms that extract insights from massive amounts of data. Therefore, such systems need to process ever-increasing amounts of data, demanding higher memory bandwidth and capacity. However, the bandwidth between processors and off-chip memory has not increased due to various stringent physical constraints. Besides, data transfers between the processors and the off-chip memory consume orders of magnitude more energy than on-chip computation due to the disparity between interconnection and transistor scaling.Exploiting recent 3D-stacking technology, the researcher community has explored near-data processing architectures that place processors and memory on the same chip. However, it is unclear whether or not such processing-in-memory (PIM) attempts will be successful for commodity computing systems due to the high cost of 3D-stacking technology and demanded change in existing processor, memory and/or applications. Faced with these challenges, the PIs are to investigate near-data processing platforms that do not require any change in processor, memory and applications, exploiting deep insights on commodity memory subsystems and network software stack. The success of this project will produce inexpensive but powerful near-data processing platforms that can directly run existing machine learning applications without any modification.

越来越多的商业和企业系统越来越依赖机器学习算法。这种转变一方面是由于机器学习算法的突破，这些算法从大量数据中提取见解。因此，这样的系统需要处理不断增加的数据量，要求更高的存储器带宽和容量。然而，由于各种严格的物理约束，处理器和片外存储器之间的带宽并没有增加。此外，由于互连和晶体管缩放之间的差异，处理器和片外存储器之间的数据传输消耗的能量比片上计算多几个数量级。利用最新的3D堆叠技术，研究人员已经探索了将处理器和存储器放置在同一芯片上的近数据处理架构。然而，由于3D堆叠技术的高成本以及现有处理器、存储器和/或应用程序中的所需改变，不清楚这种存储器中处理（PIM）尝试对于商用计算系统是否会成功。面对这些挑战，PI将研究不需要对处理器、内存和应用程序进行任何更改的近数据处理平台，并利用对商品内存子系统和网络软件堆栈的深刻见解。该项目的成功将产生廉价但功能强大的近数据处理平台，可以直接运行现有的机器学习应用程序而无需任何修改。

项目成果

期刊论文数量（8）

专著数量（0）

科研奖励数量（0）

会议论文数量（0）

专利数量（0）

GANAX: A Unified MIMD-SIMD Acceleration for Generative Adversarial Networks

DOI：
10.1109/isca.2018.00060
发表时间：
2018-05
期刊：
2018 ACM/IEEE 45th Annual International Symposium on Computer Architecture (ISCA)
影响因子：
0
作者：
A. Yazdanbakhsh;Hajar Falahati;Philip J. Wolfe;K. Samadi;N. Kim;H. Esmaeilzadeh
通讯作者：
A. Yazdanbakhsh;Hajar Falahati;Philip J. Wolfe;K. Samadi;N. Kim;H. Esmaeilzadeh

In-RDBMS Hardware Acceleration of Advanced Analytics

DOI：
10.14778/3236187.3236188
发表时间：
2018-01
期刊：
Proc. VLDB Endow.
影响因子：
0
作者：
Divya Mahajan;J. Kim;Jacob Sacks;A. Ardalan;Arun Kumar;H. Esmaeilzadeh
通讯作者：
Divya Mahajan;J. Kim;Jacob Sacks;A. Ardalan;Arun Kumar;H. Esmaeilzadeh

ReLeQ: An Automatic Reinforcement Learning Approach for Deep Quantization of Neural Networks

DOI：
发表时间：
2018
期刊：
影响因子：
0
作者：
Ahmed T. Elthakeb;Prannoy Pilligundla;FatemehSadat Mireshghallah;A. Yazdanbakhsh;H. Esmaeilzadeh
通讯作者：
Ahmed T. Elthakeb;Prannoy Pilligundla;FatemehSadat Mireshghallah;A. Yazdanbakhsh;H. Esmaeilzadeh

Divide and Conquer: Leveraging Intermediate Feature Representations for Quantized Training of Neural Networks

DOI：
发表时间：
2019-06
期刊：
ArXiv
影响因子：
0
作者：
Ahmed T. Elthakeb;Prannoy Pilligundla;H. Esmaeilzadeh
通讯作者：
Ahmed T. Elthakeb;Prannoy Pilligundla;H. Esmaeilzadeh

From Tensors to FPGAs: Accelerating Deep Learning

DOI：
发表时间：
2018-08
期刊：
影响因子：
0
作者：
Hardik Sharma;Jongse Park;Balavinayagam Samynathan;Behnam Robatmili;S. Mirkhani;H. Esmaeilzadeh
通讯作者：
Hardik Sharma;Jongse Park;Balavinayagam Samynathan;Behnam Robatmili;S. Mirkhani;H. Esmaeilzadeh