A low-complexity bit-efficient Neuromorphic Astrocyte-Neuron Circuit

Abstract

Experimental and computational evidence accumulated over the last few decades suggests that astrocytes should be requisite and active constituents in neuronal information processing by providing appropriate feedback control to regulate neuronal activities in the central nervous system (CNS). Consequently, they have an indispensable role in the brain. Bearing that in mind, the hardware realization of astrocytes is important for developing the next generation of bio-inspired computing systems. In the proposed digital astrocyte-neuron circuit, the calcium dynamics of the astrocyte are represented by a linear model, and the neuron dynamics are described by a modified version of the Izhikevich model. The results obtained from an FPGA implementation and simulation datasets, illustrate that the suggested circuit is biologically plausible, less complex and bit-efficient compared to previous competitive implementations. Compared to recent competitive implementations, the proposed neuromorphic system presents better characteristics in terms of area and performance. © 2021 IEEE.