Analogical and Relational Reasoning with Spiking Neural Networks

TL;DR

This paper demonstrates that neural networks with biologically inspired spiking modules excel at solving Raven's Progressive Matrices, surpassing human-level accuracy in supervised learning and outperforming existing methods in unsupervised learning, highlighting their potential for abstract reasoning.

Contribution

The study introduces and evaluates spiking neural modules within neural networks, showing their advantages in abstract reasoning tasks over traditional models.

Findings

Supervised spiking networks surpass human-level accuracy on RAVEN dataset.

Unsupervised spiking networks outperform existing unsupervised methods.

Spiking modules enable networks to encode temporal features and generalize better.

Abstract

Raven's Progressive Matrices have been widely used for measuring abstract reasoning and intelligence in humans. However for artificial learning systems, abstract reasoning remains a challenging problem. In this paper we investigate how neural networks augmented with biologically inspired spiking modules gain a significant advantage in solving this problem. To illustrate this, we first investigate the performance of our networks with supervised learning, then with unsupervised learning. Experiments on the RAVEN dataset show that the overall accuracy of our supervised networks surpass human-level performance, while our unsupervised networks significantly outperform existing unsupervised methods. Finally, our results from both supervised and unsupervised learning illustrate that, unlike their non-augmented counterparts, networks with spiking modules are able to extract and encode temporal features without any explicit instruction, do not heavily rely on training data, and generalise more readily to new problems. In summary, the results reported here indicate that artificial neural networks with spiking modules are well suited to solving abstract reasoning.