Neural-Inspired Dendritic Multiplication Using a Reconfigurable Analog Integrated Circuit
Jordan Edwards, Luke Parker, Suma Cardwell, Frances S. Chance, Scott Koziol · 2024
A longstanding goal of neuromorphic computing is to develop novel computer architectures with a brain-like energy footprint. Still, key biological components such as dendrites are often overlooked. Instead, more compact scalable neuron models are favored for neuromorphic implementations. Here we present a novel neuroscience-inspired multiplicative circuit that uses shunting inhibition as the multiplicative mechanism in a CMOS dendrite circuit. We demonstrate an experimental implementation of our model on a 350-nm process field programmable analog array (FPAA) integrated circuit (IC), that exhibits shunting inhibition as a multiplicative process in neuromorphic circuits.