A Framework for Neural Network Pruning Using Gibbs Distributions

TL;DR

This paper introduces Gibbs pruning, a novel framework combining statistical physics and stochastic regularization to train and prune neural networks simultaneously, leading to improved performance and state-of-the-art results on CIFAR-10.

Contribution

The paper presents Gibbs pruning, a new unified framework for neural network pruning that outperforms existing methods and achieves state-of-the-art results.

Findings

Gibbs pruning outperforms contemporary pruning methods.

Achieved state-of-the-art pruning results for ResNet-56 on CIFAR-10.

Supports both structured and unstructured pruning.

Abstract

Modern deep neural networks are often too large to use in many practical scenarios. Neural network pruning is an important technique for reducing the size of such models and accelerating inference. Gibbs pruning is a novel framework for expressing and designing neural network pruning methods. Combining approaches from statistical physics and stochastic regularization methods, it can train and prune a network simultaneously in such a way that the learned weights and pruning mask are well-adapted for each other. It can be used for structured or unstructured pruning and we propose a number of specific methods for each. We compare our proposed methods to a number of contemporary neural network pruning methods and find that Gibbs pruning outperforms them. In particular, we achieve a new state-of-the-art result for pruning ResNet-56 with the CIFAR-10 dataset.