Spikes That Count: Rethinking Spikiness In Neurally Embedded Systems

Spikes That Count: Rethinking Spikiness In Neurally Embedded Systems

Abstract

Spiky neural networks are widely used in neural modeling, due to their biological rel- evance and high computational power. In this paper we investigate the usage of spiking dynamics in embedded arti¯cial neural networks, that serve as a control mechanism for evolved autonomous agents performing a delayed-response task. The synaptic weights and spiking dynamics are evolved using a genetic algorithm. We compare evolved spiky networks with evolved McCulloch-Pitts networks, while confronting new questions about the nature of \spikiness" and its contribution to the neurocontroller's processing. On the behavioral level, we show that in a memory-dependent task, network solutions that incorporate spiking dynamics can be less complex and easier to evolve than neurocon- trollers involving McCulloch-Pitts neurons. On the functional level, we identify and rigorously characterize two distinct properties of spiking dynamics in embedded agents: spikiness evident in°uence and spikiness functional contribution.