Unveiling Invariances via Neural Network Pruning

TL;DR

This paper introduces a pruning-based framework to learn neural network architectures that automatically capture data-specific invariances, leading to improved performance and efficiency across vision and tabular datasets.

Contribution

It presents a novel method for architecture learning through pruning to encode invariances, surpassing traditional handcrafted invariance handling.

Findings

Learned architectures outperform dense networks in accuracy.

Pruned models are more efficient in computation.

Framework is effective across diverse datasets.

Abstract

Invariance describes transformations that do not alter data's underlying semantics. Neural networks that preserve natural invariance capture good inductive biases and achieve superior performance. Hence, modern networks are handcrafted to handle well-known invariances (ex. translations). We propose a framework to learn novel network architectures that capture data-dependent invariances via pruning. Our learned architectures consistently outperform dense neural networks on both vision and tabular datasets in both efficiency and effectiveness. We demonstrate our framework on multiple deep learning models across 3 vision and 40 tabular datasets.