Improving deep neural network random initialization through neuronal rewiring

Author(s)

Scabini, Leonardo ; De Baets, Bernard ; Bruno, Odemir M.

• Document type: Journal article
• Source: Neurocomputing; v. 599, p. 13-pg., 2024-07-17.

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 on image classification tasks 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. (AU)

• FAPESP's process: 19/07811-0 - Artificial neural networks and complex networks: an integrative study of topological properties and pattern recognition
• Grantee: Leonardo Felipe dos Santos Scabini
• Support Opportunities: Scholarships in Brazil - Doctorate
• FAPESP's process: 21/09163-6 - Network science for optimizing artificial neural networks on computer vision
• Grantee: Leonardo Felipe dos Santos Scabini
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• FAPESP's process: 21/08325-2 - An analysis of network automata as models for biological and natural processes
• Grantee: Odemir Martinez Bruno
• Support Opportunities: Regular Research Grants
• FAPESP's process: 18/22214-6 - Towards a convergence of technologies: from sensing and biosensing to information visualization and machine learning for data analysis in clinical diagnosis
• Grantee: Osvaldo Novais de Oliveira Junior
• Support Opportunities: Research Projects - Thematic Grants