Information Dimension Matching in Memristive Computing System for Analog Deployment of Deep Neural Networks
Zhe Feng,Zuheng Wu,Xu Wang,Xiuquan Fang,Xumeng Zhang,Jianxun Zou,Jian Lu,Wenbin Guo,Xing Li,Tuo Shi,Zuyu Xu,Yunlai Zhu,Fei Yang,Yuehua Dai,Qi Liu
DOI: https://doi.org/10.1002/aelm.202400106
IF: 6.2
2024-05-30
Advanced Electronic Materials
Abstract:This study introduces the Information Dimension Matching (IDM) scheme for deploying Deep Neural Networks (DNNs) into Memristive Computing Systems (MCS). IDM synchronously scales input voltage and target conductance during matrix‐vector multiplication to match information dimensions. The IDM significantly improves the inference accuracy and energy efficiency of MCS for DNN deployment. Memristor, with the ability of analog computing, is widely investigated for improving the computing efficiency of deep neural networks (DNNs) deployment. However, how to fully take advantage of the analog computing ability of memristive computing system (MCS) for DNN deployment is still an open question. Here, a new neural network models deployment scheme, that is, an information dimension matching (IDM) scheme, is proposed to fully take advantage of the analog computing ability of MCS. Furthermore, the spatial and temporal DNN, that is convolutional neural network (CNN) and recurrent neural network (RNN) is used to verify the proposed deployment scheme, respectively. The experimental results indicate that, compared to the traditional deployment schemes, the proposed deployment scheme shows obvious inference accuracy and energy efficiency improvement (>4 × in four‐layer DNNs deployment), and the energy efficiency improvement increases dramatically with the layers increment of DNNs. This work paves the path for developing high computing efficiency analog MCS.
materials science, multidisciplinary,physics, applied,nanoscience & nanotechnology