Photonic classification on a single diffractive layer

Abstract

Photonic computation started to shape the future of fast, efficient and accessible computation. The advantages brought by light based Diffractive Deep Neural Networks (D2NN), are shown to be overwhelmingly advantageous especially in targeting classification problems. However, cost and complexity of multi-layer systems are the main challenges that reduce the deployment of this technology. In this study, we develop a simple yet extremely efficient way to achieve optical classification using a single diffractive optical layer. A spatial light modulator is used not only to emulate the classifying system but also the input medium. We show that using a simple interpretable linear classifier, images can be classified at the speed of light. We perform classification of road signs under the effect of noise and demonstrate that we can successfully classify input images with more than 90% accuracy even with 13% noise/imperfection.