FPGA Simulation of Spiking Neuron Signals

Abstract

Recently, with the increasing demand for artificial intelligence applications, there is a growing need for neuromorphic structures that most closely resemble the living brain. Based on these structures, Spiking Neural Networks, defined as third-generation Artificial Neural Networks using neurological signals, are being deployed. As an alternative to traditional computer architectures, neuromorphic hardware designed specifically for artificial intelligence offers advantages in power consumption, speed, and resource usage. The IEEE754 32-bit number format provides flexibility in the binary representation of decimals compared to fixed-point numbers. In this study, the neurological signals which are mathematically modeled by Izhikevich were first determined in the Matlab environment. Signals with the same pattern were simulated with single precision in the IEEE754 standard of floating-point arithmetic on the ISE Design Suite environment for use in future studies on FPGA hardware. While the studies in the literature show fixed-point arithmetic, this study presents the structure that should be used for floating-point arithmetic.