Abstract
AbstractThis is a preliminary thesis on Temporal Difference Spike-Timing Dependent Plasticity (TDSTDP), a variation of STDP that considers dendritic potential. TDSTDP is capable of performing temporal difference (TD) learning without dopamine modulation. Major characteristics of TD learning, including value estimation, value propagation, and temporal shifting of dopamine, are demonstrated in simulations. Additionally, a synaptic calcium model demonstrates its biological plausibility.
Publisher
Cold Spring Harbor Laboratory