Improving Deep Neural Network Random Initialization Through Neuronal Rewiring

TL;DR

This paper introduces a neuronal rewiring method based on preferential attachment to improve weight initialization in deep neural networks, leading to better performance by reducing neuronal strength variance.

Contribution

It proposes a novel rewiring technique that reorganizes connections based on neuronal strength, enhancing initialization without altering weight magnitudes.

Findings

Rewiring reduces neuronal strength variance.

Performance improves during training and testing.

Effective across various architectures and schedules.

Abstract

The deep learning literature is continuously updated with new architectures and training techniques. However, weight initialization is overlooked by most recent research, despite some intriguing findings regarding random weights. On the other hand, recent works have been approaching Network Science to understand the structure and dynamics of Artificial Neural Networks (ANNs) after training. Therefore, in this work, we analyze the centrality of neurons in randomly initialized networks. We show that a higher neuronal strength variance may decrease performance, while a lower neuronal strength variance usually improves it. A new method is then proposed to rewire neuronal connections according to a preferential attachment (PA) rule based on their strength, which significantly reduces the strength variance of layers initialized by common methods. In this sense, PA rewiring only reorganizes connections, while preserving the magnitude and distribution of the weights. We show through an extensive statistical analysis in image classification that performance is improved in most cases, both during training and testing, when using both simple and complex architectures and learning schedules. Our results show that, aside from the magnitude, the organization of the weights is also relevant for better initialization of deep ANNs.