Perceptrons with Hebbian learning based on wave ensembles in plastic potentials

We present a general theoretical model to realize a bilayer perceptron for hardware neural networks with applications in pattern recognition. In the network, multiple interconnections are allowed, by using the Schrödinger wave function as input and outputs signals; moreover, microscopic plastic potentials allow to process information and “train” the system in micrometer’s scale. As particular cases, we present the calculations for two devices where the information is carried by light and the plastic potentials emerge due polariton-phonon and polariton-nuclear spin interactions. We show both designs are capable to perform digit recognition.