Graphene-Based Photonic-Electronic Multiply-Accumulate Neurons
Lorenzo De Marinis, P. S. Kincaid, G. Contestabile, Sumanta Gupta, Nicola Andriolli · 2023
We assess the potential of graphene optical modulators for high-speed neuromorphic computing. We numerically validate the performance in terms of analog resolution of a photonic-electronic neuron based on high-speed graphene electro-absorption modulators. The neuromorphic processor delivers up to 280 GMAC/s with a resolution spanning from 2.7 to 5.4 equivalent bits for different operating frequencies.