Implicit Filter Sparsification In Convolutional Neural Networks

TL;DR

This paper investigates how implicit filter sparsity naturally occurs in CNNs with Batch Normalization and ReLU, revealing mechanisms that can inform filter pruning strategies.

Contribution

It provides an empirical analysis of implicit filter sparsification in CNNs, linking it to existing heuristics and suggesting new pruning approaches.

Findings

Implicit sparsity emerges in CNN filters during training.

Selective feature pruning contributes to sparsity.

Implications for filter pruning heuristics.

Abstract

We show implicit filter level sparsity manifests in convolutional neural networks (CNNs) which employ Batch Normalization and ReLU activation, and are trained with adaptive gradient descent techniques and L2 regularization or weight decay. Through an extensive empirical study (Mehta et al., 2019) we hypothesize the mechanism behind the sparsification process, and find surprising links to certain filter sparsification heuristics proposed in literature. Emergence of, and the subsequent pruning of selective features is observed to be one of the contributing mechanisms, leading to feature sparsity at par or better than certain explicit sparsification / pruning approaches. In this workshop article we summarize our findings, and point out corollaries of selective-featurepenalization which could also be employed as heuristics for filter pruning