Neural networks based on ultrafast time-delayed effects in exciton-polaritons

TL;DR

This paper presents a novel neural network model utilizing ultrafast time-delayed nonlinear effects in exciton-polaritons, enabling high-speed optical information processing and classification tasks.

Contribution

It introduces a new approach to neural networks based on exciton-polariton nonlinearities and demonstrates a functional XOR gate and digit classification at picosecond timescales.

Findings

Successful implementation of a nonlinear XOR logic gate.

High-accuracy classification of spoken digits.

Operation on picosecond timescale.

Abstract

We demonstrate that time-delayed nonlinear effects in exciton-polaritons can be used to construct neural networks where information is coded in optical pulses arriving consecutively on the sample. The highly nonlinear effects are induced by time-dependent interactions with the excitonic reservoir. These nonlinearities allow to create a nonlinear XOR logic gate that can perform operations on the picosecond timescale. An optoelectronic neural network based on the constructed logic gate performs classification of spoken digits with a high accuracy rate.