Dynamic Memory via Delay Coincidence Detection for Robot Maze Navigation


This paper demonstrates that spatiotemporal patterns produced by spiking neural networks of coincidence detection cells can be effectively decoded. We begin by recounting a procedure to find appropriate decoding delays. We expand on previous work by showing that an accurate coincidence detection threshold for decoding cells can be determined from observed dynamics. We then show that randomly generated networks can serve as dynamic memories in a two-turn embodied agent maze task. These results supports the claim that biological neural circuits of coincidence detection cells play a potentially significant role in cognition via time based coding.

Back to Table of Contents