# Locality-Promoting Representation Learning

## TL;DR

This paper reveals that CNN filters tend to have larger weights near the center, introduces a regularization method to promote this locality, and demonstrates improved accuracy across various architectures and datasets.

## Contribution

The paper introduces Locality-promoting Regularization (LOCO-Reg) to enforce spatial locality in CNN filters, improving performance and providing theoretical insights.

## Key findings

- Weights near filter centers are larger than those on the outside.
- LOCO-Reg improves accuracy across multiple CNN architectures.
- The empirical locality pattern is explained by maximizing feature cohesion.

## Abstract

This work investigates fundamental questions related to learning features in convolutional neural networks (CNN). Empirical findings across multiple architectures such as VGG, ResNet, Inception, DenseNet and MobileNet indicate that weights near the center of a filter are larger than weights on the outside. Current regularization schemes violate this principle. Thus, we introduce Locality-promoting Regularization (LOCO-Reg), which yields accuracy gains across multiple architectures and datasets. We also show theoretically that the empirical finding is a consequence of maximizing feature cohesion under the assumption of spatial locality.