Evolving spiking neural networks: A novel growth algorithm corrects the teacher

Abstract

Spiking neural networks (SNNs) have generated considerable excitement because of their computational properties, believed to be superior to conventional von Neumann machines, and sharing properties with living brains. Yet progress building these systems has been limited because we lack a design methodology. We present a gene-driven network growth algorithm that enables a genetic algorithm (evolutionary computation) to generate and test SNNs. The genome length for this algorithm grows O(n) where n is the number of neurons; n is also evolved. The genome not only specifies the network topology, but all its parameters as well. In experiments, the algorithm discovered SNNs that effectively produce a robust spike bursting behavior given tonic inputs, an application suitable for central pattern generators. Even though evolution did not include perturbations of the input spike trains, the evolved networks showed remarkable robustness to such perturbations. On a second task, a sequence detector, several related discriminating designs were found, all made “errors” in that they fired when input spikes were simultaneous (i.e. not strictly in sequence), but not when they were out of sequence. They also fired when the sequence was too close for the teacher to have declared they were in sequence. That is, evolution produced these behaviors even though it was not explicitly rewarded for doing so. We are optimistic that this technology might be scaled up to produce robust SNN designs that humans would be hard pressed to produce.