期刊
IEEE TRANSACTIONS ON COMPUTER-AIDED DESIGN OF INTEGRATED CIRCUITS AND SYSTEMS
卷 40, 期 5, 页码 971-984出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TCAD.2020.3013194
关键词
3-D; convolutional neural network (CNN); heterogeneous; resistive random-access memory (ReRAM); thermal
类别
资金
- U.S. National Science Foundation (NSF) [CNS-1564014, CCF 1514269, CSR-1717885]
- USA Army Research Office [W911NF-17-1-0485]
The paper introduces an M3D-enabled heterogeneous architecture, AccuReD, that combines ReRAM arrays with GPU cores to address challenges in CNN training and achieve high accuracy. AccuReD supports all types of CNN layers and achieves near-GPU accuracy even with the low-precision and nonideal behavior of ReRAMs, accelerating CNN training while maintaining accuracy.
The growing popularity of convolutional neural networks (CNNs) along with their complexity has led to the search for efficient computational platforms suitable for them. Resistive random-access memory (ReRAM)-based architectures offer a promising alternative to commonly used GPU-based platforms for training CNNs. However, due to their low-precision storage capability, these architectures cannot support all types of CNN layers and suffer from accuracy loss of the learned model. In addition, ReRAM behavior varies with temperature. High temperature reduces noise margin and introduces additional noise. This makes training of CNNs challenging as outputs can be misinterpreted at higher operating temperatures leading to accuracy loss. In this work, we propose an M3D-enabled heterogeneous architecture: AccuReD, that combines ReRAM arrays with GPU cores, to address these challenges and achieve high accuracy CNN training. AccuReD supports all types of CNN layers and achieve near-GPU accuracy even with low-precision and nonideal behavior of ReRAMs. In addition, to reduce temperature, we present a performance-thermal-aware mapping policy that maps CNN layers to the computing elements of AccuReD. Experimental evaluation indicates that AccuReD does not lose accuracy while accelerating CNN training by 12x on an average compared to conventional GPU-only platforms.
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