Thomas Robert Newton

FORCE trained spiking networks do not benefit from faster learning while parameter matched rate networks do

This paper compares training spiking and rate-based neural networks using FORCE, finding that slower learning rates stabilize performance in spiking networks while faster learning causes instability and uncorrelated solutions.

Does the Zebra Finch Mating Song Circuit Use Spike Times Efficiently?

The paper shows that precise and reliable spike timing in neural circuits improves encoding accuracy more efficiently than rate coding, particularly in larger networks.