Learning Sparse Neural Networks with Identity Layers

TL;DR

This paper explores the relationship between interlayer feature similarity and network sparsity, proposing a CKA-based regularization method that enhances sparsity and performance in neural networks.

Contribution

It introduces a novel CKA-based regularization technique that reduces feature similarity between layers, promoting sparsity and improving existing sparse training methods.

Findings

CKA-SR improves sparsity in neural networks.

Reducing feature similarity enhances network performance.

Method is effective at extremely high sparsity levels.

Abstract

The sparsity of Deep Neural Networks is well investigated to maximize the performance and reduce the size of overparameterized networks as possible. Existing methods focus on pruning parameters in the training process by using thresholds and metrics. Meanwhile, feature similarity between different layers has not been discussed sufficiently before, which could be rigorously proved to be highly correlated to the network sparsity in this paper. Inspired by interlayer feature similarity in overparameterized models, we investigate the intrinsic link between network sparsity and interlayer feature similarity. Specifically, we prove that reducing interlayer feature similarity based on Centered Kernel Alignment (CKA) improves the sparsity of the network by using information bottleneck theory. Applying such theory, we propose a plug-and-play CKA-based Sparsity Regularization for sparse network training, dubbed CKA-SR, which utilizes CKA to reduce feature similarity between layers and increase network sparsity. In other words, layers of our sparse network tend to have their own identity compared to each other. Experimentally, we plug the proposed CKA-SR into the training process of sparse network training methods and find that CKA-SR consistently improves the performance of several State-Of-The-Art sparse training methods, especially at extremely high sparsity. Code is included in the supplementary materials.