Co-design of DNN model optimization for binary ReRAM array in-memory processing

Yue Guan, Takashi Ohsawa

Research output: Chapter in Book/Report/Conference proceedingConference contribution

2 Citations (Scopus)

Abstract

In this work, a novel design of ReRAM neuromorphic system is proposed to process deep neural network (DNN) fully in array efficiently. A binary neural network model is constructed and optimized on MNIST dataset. The obtained model is simulated to be processed with the proposed ReRAM array. Co-design between hardware and network model optimization in software is analyzed to achieve feasible hardware design and generalizable model. Deployed with such co-design model, ReRAM array processes DNN with high robustness against fabrication fluctuation.

Original languageEnglish
Title of host publication2019 IEEE 11th International Memory Workshop, IMW 2019
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781728109817
DOIs
Publication statusPublished - 2019 May
Event11th IEEE International Memory Workshop, IMW 2019 - Montterey, United States
Duration: 2019 May 122019 May 15

Publication series

Name2019 IEEE 11th International Memory Workshop, IMW 2019

Conference

Conference11th IEEE International Memory Workshop, IMW 2019
CountryUnited States
CityMontterey
Period19/5/1219/5/15

Keywords

  • binary neural network
  • fabrication fluctuation
  • neuromorphic ReRAM

ASJC Scopus subject areas

  • Hardware and Architecture
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Co-design of DNN model optimization for binary ReRAM array in-memory processing'. Together they form a unique fingerprint.

  • Cite this

    Guan, Y., & Ohsawa, T. (2019). Co-design of DNN model optimization for binary ReRAM array in-memory processing. In 2019 IEEE 11th International Memory Workshop, IMW 2019 [8739722] (2019 IEEE 11th International Memory Workshop, IMW 2019). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/IMW.2019.8739722