Topology and Computational Performance of Attractor Neural Networks

TL;DR

This paper investigates how different network topologies, including regular, random, small-world, and scale-free, influence the computational efficiency and robustness of attractor neural networks, revealing unique advantages of scale-free structures.

Contribution

It provides a comparative analysis of various network topologies on neural network performance, highlighting the robustness of scale-free networks in pattern recognition.

Findings

Random networks are most efficient for pattern storage and retrieval.

Scale-free networks exhibit robust partial recognition focused on highly connected nodes.

Network topology significantly affects neural network computational performance.

Abstract

To explore the relation between network structure and function, we studied the computational performance of Hopfield-type attractor neural nets with regular lattice, random, small-world and scale-free topologies. The random net is the most efficient for storage and retrieval of patterns by the entire network. However, in the scale-free case retrieval errors are not distributed uniformly: the portion of a pattern encoded by the subset of highly connected nodes is more robust and efficiently recognized than the rest of the pattern. The scale-free network thus achieves a very strong partial recognition. Implications for brain function and social dynamics are suggestive.