跳到主要导航 跳到搜索 跳到主要内容

Approximate computation based on NAND-SPIN MRAM for CNN on-chip training

  • Beihang University

科研成果: 书/报告/会议事项章节会议稿件同行评审

摘要

Approximate computation is a widely used method to accelerate CNN training. In this work, the stochastic switching mechanism of the NAND-SPIN MRAM is utilized to perform the approximate update and storage of the synaptic weight. By reducing the programming time of the NAND-SPIN MTJs from 3 ns to 1 ns, more than 67% speedup and nearly 70% energy saving have been achieved with less than 1% accuracy loss.

源语言英语
主期刊名Proceedings of the 17th ACM International Symposium on Nanoscale Architectures, NANOARCH 2022
出版商Association for Computing Machinery, Inc
ISBN(电子版)9781450399388
DOI
出版状态已出版 - 7 12月 2022
活动17th ACM International Symposium on Nanoscale Architectures, NANOARCH 2022 - Virtual, Online, 美国
期限: 7 12月 20229 12月 2022

出版系列

姓名Proceedings of the 17th ACM International Symposium on Nanoscale Architectures, NANOARCH 2022

会议

会议17th ACM International Symposium on Nanoscale Architectures, NANOARCH 2022
国家/地区美国
Virtual, Online
时期7/12/229/12/22

指纹

探究 'Approximate computation based on NAND-SPIN MRAM for CNN on-chip training' 的科研主题。它们共同构成独一无二的指纹。

引用此