FPGA Implementation of Bio-inspired Computing Based Deep Learning Model

Abstract

Computer architecture classes are subjected to power, thermal, and area limitations and suggest facilitating optimization of chip performance for a specific device technology through an optimal performance, memory subsystems, and interconnection model for various technology architectures. The goal of Bio-Inspired engineering is to use Very Large Scale Integration (VLSI) as well as analog electronic circuits to reverse engineer the human brain. This research manuscript presents brain-inspired computing architecture based on the Field Programmable Gate Array (FPGA) neuron model Leaky-Integrate-and-File (LIF) for the massively parallel bio-inspired computing system. In designing a field-programmable neuromorphic computing system, the reconfigurable and event-driven parameters are considered. In this paper an event driven FPGA system is presented which is the combination of memory arbiter for logical memory access for kernel processing. The Register Transfer Logic (RTL) results of implementation and hardware synthesis are presented as a proof of concept. The neuron model is explored in Xilinx-ISE software with utilizing VERILOG code, considering digital implementation, aiming at high-speed low-cost large-scale systems. The deep learning algorithms interact with the FPGA model developed and we carry out the simulations of neuron for the membrane potential in the corresponding receptive field, along with obtaining the characteristics plot between membrane potential and the receptive field.