CogniGron Seminar: Marcelo Rozenberg (CNRS - Université Paris Saclay) - "Solid State Neuroscience"

In this talk I shall introduce some basic notions of electronic neuromorphic functionalities of materials and devices. I shall first briefly describe our efforts towards implementing artificial neurons exploiting the electric Mott transition. Then, I shall describe in more detail our recent work implementing a new silicon-based memristive neuron. This artificial spiking neuron is extremely simple as it exploits an overlooked memristive feature of a “forgotten” conventional electronic device, the thyristor. The main features of the memristive silicon neuron circuit are that it provides a simple and physical spiking neuron model. As a counterpart of the Izhikevich neuron mathematical model, it can be easily adapted to capture multiple biologically relevant spiking patterns. More importantly, this neuron circuit can be associated as LEGO blocks to build spiking neural networks that are relevant to Neuroscience, which we are beginning to explore.