Bio-inspired metrics for synthetic neuromorphic systems

Neuromorphic computing proposes an alternative to traditional computing by developing hardware that integrates and processes information through mechanisms analogous to biological neurons.The main motivations for this research are to develop more compatible hardware for artificial neural networks, and to to study diverse topics of neuroscience. Such hardware requires properties compatible with brain-inspired computation, such as spiking behavior, network dynamics, scale-free morphology, and many others. The goal of this work is to develop metrics that quantify the resemblance of these neuromorphic systems to biological neurons. We present the results of this analysis in high density silver nanowires networks, which are promising material for neuromorphic applications. First, we characterized their morphology and identified fractal-like properties applying a box-counting algorithm to optical images. Second, we observed spike-like electrical responses to external stimuli, which were compared with neuronal activity generated by the Izhikevich model. We are currently extending this analysis by studying a biological computer developed by Cortical Labs, which is a platform that runs in a culture of human neurons. Particularly, we are investigating correlation metrics, focusing on the spike time tiling coefficient proposed by Cutts and Eglen. With these metrics we aim to provide a framework for comparing the neuromorphic potential of diverse biological and artificial systems.