Programmable neuromorphic circuits for spike-based neural dynamics

Abstract

Hardware implementations of spiking neural networks offer promising solutions for a wide set of tasks, ranging from autonomous robotics to brain machine interfaces. We propose a set of programmable hybrid analog/digital neuromorphic circuits than can be used to build compact low-power neural processing systems. In particular, we present both CMOS and hybrid memristor/CMOS synaptic circuits that have programmable synaptic weights and exhibit biologically plausible response properties. For the CMOS circuits, we present experimental results demonstrating that they operate correctly over a wide range input frequencies; for the hybrid memristor/CMOS circuits we present circuit simulation results validating their expected response properties.