Sifting out the features by pruning: Are convolutional networks the winning lottery ticket of fully connected ones?

TL;DR

This paper investigates how pruning creates sub-networks in fully connected neural networks that resemble convolutional networks, revealing that pruning can uncover CNN-like features and architectures.

Contribution

The study demonstrates that iterative pruning of fully connected networks results in architectures with local connectivity patterns similar to CNNs, highlighting the role of pruning in discovering inductive biases.

Findings

Pruned FCNs develop local connectivity patterns akin to CNNs.

Winning lottery tickets in FCNs exhibit features characteristic of CNNs.

Pruning can uncover architectures with CNN-like inductive biases.

Abstract

Pruning methods can considerably reduce the size of artificial neural networks without harming their performance. In some cases, they can even uncover sub-networks that, when trained in isolation, match or surpass the test accuracy of their dense counterparts. Here we study the inductive bias that pruning imprints in such "winning lottery tickets". Focusing on visual tasks, we analyze the architecture resulting from iterative magnitude pruning of a simple fully connected network (FCN). We show that the surviving node connectivity is local in input space, and organized in patterns reminiscent of the ones found in convolutional networks (CNN). We investigate the role played by data and tasks in shaping the architecture of pruned sub-networks. Our results show that the winning lottery tickets of FCNs display the key features of CNNs. The ability of such automatic network-simplifying procedure to recover the key features "hand-crafted" in the design of CNNs suggests interesting applications to other datasets and tasks, in order to discover new and efficient architectural inductive biases.