Permute, Quantize, and Fine-tune: Efficient Compression of Neural Networks

TL;DR

This paper introduces a permutation-based approach combined with quantization and fine-tuning to improve neural network compression, especially for modern architectures, achieving significant accuracy retention across vision tasks.

Contribution

It proposes a novel permutation strategy for weights that enhances vector quantization efficiency, connecting to rate-distortion theory for better compression.

Findings

Achieves 40-70% gap reduction with uncompressed models

Improves compression for pointwise convolutions and linear layers

Enhances accuracy with annealed quantization

Abstract

Compressing large neural networks is an important step for their deployment in resource-constrained computational platforms. In this context, vector quantization is an appealing framework that expresses multiple parameters using a single code, and has recently achieved state-of-the-art network compression on a range of core vision and natural language processing tasks. Key to the success of vector quantization is deciding which parameter groups should be compressed together. Previous work has relied on heuristics that group the spatial dimension of individual convolutional filters, but a general solution remains unaddressed. This is desirable for pointwise convolutions (which dominate modern architectures), linear layers (which have no notion of spatial dimension), and convolutions (when more than one filter is compressed to the same codeword). In this paper we make the observation that the weights of two adjacent layers can be permuted while expressing the same function. We then establish a connection to rate-distortion theory and search for permutations that result in networks that are easier to compress. Finally, we rely on an annealed quantization algorithm to better compress the network and achieve higher final accuracy. We show results on image classification, object detection, and segmentation, reducing the gap with the uncompressed model by 40 to 70% with respect to the current state of the art.