Quantized neural network design under weight capacity constraint

TL;DR

This paper evaluates the trade-offs between network size scaling and weight quantization in neural networks for hardware efficiency, introducing the effective compression ratio to optimize resource use.

Contribution

It provides an analysis of neural network performance under different complexity and weight precision constraints, proposing the effective compression ratio as a new metric.

Findings

Quantization impacts neural network performance significantly.

Network size scaling can be more effective than weight quantization under certain conditions.

The effective compression ratio guides hardware-efficient neural network design.

Abstract

The complexity of deep neural network algorithms for hardware implementation can be lowered either by scaling the number of units or reducing the word-length of weights. Both approaches, however, can accompany the performance degradation although many types of research are conducted to relieve this problem. Thus, it is an important question which one, between the network size scaling and the weight quantization, is more effective for hardware optimization. For this study, the performances of fully-connected deep neural networks (FCDNNs) and convolutional neural networks (CNNs) are evaluated while changing the network complexity and the word-length of weights. Based on these experiments, we present the effective compression ratio (ECR) to guide the trade-off between the network size and the precision of weights when the hardware resource is limited.