Hardware-In-The-Loop Training of a 4f Optical Correlator with Logarithmic Complexity Reduction for CNNs

TL;DR

This paper demonstrates hardware-in-the-loop training of a 4f optical correlator for CNNs, achieving high accuracy with reduced complexity, offering a promising approach for optical neural network training.

Contribution

It introduces a forward-only learning algorithm for optical correlators with logarithmic complexity reduction, enabling efficient CNN training.

Findings

Achieved 87.6% accuracy on MNIST

Reduced complexity from O(n2 log n) to O(n2)

Demonstrated feasibility of optical hardware-in-the-loop training

Abstract

This work evaluates a forward-only learning algorithm on the MNIST dataset with hardware-in-the-loop training of a 4f optical correlator, achieving 87.6% accuracy with O(n2) complexity, compared to backpropagation, which achieves 88.8% accuracy with O(n2 log n) complexity.