Exploring Spiking Neuron Model behaviours through the Analysis of Parameter Space

As the field of deep learning rapidly gains prominence for IoT edge devices, researchers are driven to explore approaches that emulate the remarkable ability and efficiency of the human brain in solving complex tasks. Within the domain of neuromorphics, the objective is to replicate the capabilities and efficiency of the human brain and its fundamental processing unit, the neuron. This is achieved by developing models that can capture the behaviors of the latter, with various methodologies employed to study them by means of mathematical tools such as nonlinear dynamics through phase space or numerical solutions via simulation. In this study, we propose an alternative approach called the Neuronal Phase Map (NePhaM), which provides a graphical representation of the neuro-computational features exhibited by spiking neuron models as parameters vary. It offers an interpretable graphical representation of the neural model's behavior while attempting to overcome the challenges associated with representing high-dimensional systems of equations. By introducing NePhaM, this work presents a novel and intuitive tool for visualizing and analyzing the computational properties of spiking neuron models across a range of parameter values. This approach aims to facilitate a deeper understanding of neural dynamics as well as to provide a tool for the design and prototyping of neuro-inspired solutions. It contributes to the ongoing efforts in developing brain-inspired models to tackle complex tasks through neuromorphic computing. Especially for spike-encoding stages, NePhaM can help in interfacing SNNs with real-valued input originating from the external world.