Calendar of Events

Interdisciplinary Special Lecture:   All Welcome

Speaker:  Marcelo Rozenberg, CNRS

Abstract:  The idea of using electronic circuits to mimic the brain—known as neuromorphic electronics—dates back to Carver Mead and spans models from simple leaky-integrate-and-fire neurons to detailed Hodgkin–Huxley dynamics. Implementing spiking neural networks directly in hardware is highly attractive: it enables real-time operation without numerical approximations, scales naturally with system size, and yields neuromorphic circuits ready for practical deployment. Yet, traditional CMOS-based implementations remain complex, demand specialized fabrication facilities, and are limited by device variability.
In our work, we pursue a radically different and much simpler path based on memristive behavior. I will first describe our decade-long effort to realize artificial neurons using a remarkable property of quantum materials—the Mott metal–insulator transition. I will then turn to our more recent work on a new spiking neuron built from an inexpensive, conventional off-the-shelf electronic component: the overlooked thyristor. I will also show how equally simple synaptic circuits allow us to quantitatively reproduce textbook neuron and synapse models, enabling the implementation of basic spiking neural network motifs.
Our approach establishes a novel “theoretical hardware” framework for neuroscience, with promising applications in brain–machine interfaces, autonomous systems, and robotics.

del Valle et al; Nature (2019)
Rocco et al; Phys Rev App (2022)
Rozenberg, Schneegans and Stoliar; Scientific Reports (2019)
Wu, Wang, Schneegans, Stoliar, Rozenberg; Neuromorphic Computing and Engineering (2023)
Fernandez et al; Chaos, Solitons & Fractals (2024)
Wu, d’Hollande, Du, Rozenberg Phys Rev Applied (2025)

Host: Gabi Kotliar