WLD-Reg: A Data-dependent Within-layer Diversity Regularizer

TL;DR

This paper introduces WLD-Reg, a regularizer that promotes within-layer diversity by measuring pairwise neuron output similarities, improving neural network performance across various tasks.

Contribution

It proposes a novel within-layer diversity regularizer that complements traditional feedback, enhancing neural network training and performance.

Findings

Improves accuracy of state-of-the-art models

Enhances diversity within neural network layers

Demonstrates effectiveness across multiple tasks

Abstract

Neural networks are composed of multiple layers arranged in a hierarchical structure jointly trained with a gradient-based optimization, where the errors are back-propagated from the last layer back to the first one. At each optimization step, neurons at a given layer receive feedback from neurons belonging to higher layers of the hierarchy. In this paper, we propose to complement this traditional 'between-layer' feedback with additional 'within-layer' feedback to encourage the diversity of the activations within the same layer. To this end, we measure the pairwise similarity between the outputs of the neurons and use it to model the layer's overall diversity. We present an extensive empirical study confirming that the proposed approach enhances the performance of several state-of-the-art neural network models in multiple tasks. The code is publically available at \url{https://github.com/firasl/AAAI-23-WLD-Reg}