# Accurate deep neural network inference using computational phase-change memory

## TL;DR

This paper presents a training methodology for deep neural networks that ensures minimal accuracy loss when deploying on phase-change memory-based in-memory computing hardware, enabling energy-efficient inference.

## Contribution

It introduces a novel training approach and compensation technique for PCM-based in-memory computing, achieving high accuracy retention on CIFAR-10 and ImageNet datasets.

## Key findings

- Achieved 93.7% accuracy on CIFAR-10 after mapping to PCM
- Attained 71.6% top-1 accuracy on ImageNet with PCM hardware
- Maintained over 93.5% accuracy on CIFAR-10 over one day

## Abstract

In-memory computing is a promising non-von Neumann approach for making energy-efficient deep learning inference hardware. Crossbar arrays of resistive memory devices can be used to encode the network weights and perform efficient analog matrix-vector multiplications without intermediate movements of data. However, due to device variability and noise, the network needs to be trained in a specific way so that transferring the digitally trained weights to the analog resistive memory devices will not result in significant loss of accuracy. Here, we introduce a methodology to train ResNet-type convolutional neural networks that results in no appreciable accuracy loss when transferring weights to in-memory computing hardware based on phase-change memory (PCM). We also propose a compensation technique that exploits the batch normalization parameters to improve the accuracy retention over time. We achieve a classification accuracy of 93.7% on the CIFAR-10 dataset and a top-1 accuracy on the ImageNet benchmark of 71.6% after mapping the trained weights to PCM. Our hardware results on CIFAR-10 with ResNet-32 demonstrate an accuracy above 93.5% retained over a one day period, where each of the 361,722 synaptic weights of the network is programmed on just two PCM devices organized in a differential configuration.