LIQUID STATE MACHINE MODEL WITH HOMEOSTASIS AND SUPERVISED STDP ON NEUROMORPHIC LOIHI PROCESSOR

Abstract

This research focuses on the implementation of the Liquid State Machine model with Intrinsic Plasticity (IP) for the reservoir layer and Synaptic Plasticity for the readout layer on Intel’s new digital neuromorphic processor Loihi. Synaptic plasticity refers to the modification of weights in order to learn and infer certain patterns using a learning rule. The learning rule adopted for this model is supervised and local. Intrinsic plasticity refers to modification of neuronal states such as threshold voltage to maintain homeostasis. A Liquid State Machine Model with the combination of a homeostatic rule and a local learning rule is created on the Loihi platform and benchmarked on a speech dataset to verify its performance.