R(t)-Based Spike-Timing-Dependent Plasticity in Memristive Neural Networks

Abstract

Inspired by the human brain, neuromorphic computation should be extremely efficient at very large scales due to inherent parallelism, scalability, and fault and failure tolerance. Spike- Timing-Dependent Plasticity (STDP) is one of the most biologically plausible synaptic learning behaviors. The proposed generic model of time-varying resistance, or R(t) elements in this work can produce STDP in electronic spiking neural networks with memristive synapses that is very similar to that observed in biology. Both pair-based and triplet-based STDP is verified with the proposed generic R(t) model.