Fiber-based diffractive deep neural network

TL;DR

This paper introduces a fiber-based diffractive deep neural network that leverages waveguide mode coupling for high-performance, all-optical machine learning tasks, offering an energy-efficient alternative to electronic computing.

Contribution

It presents a novel fiber-based diffractive neural network architecture optimized for waveguide mode coupling, achieving high accuracy in complex classification tasks.

Findings

Achieved high performance in biomedical, fashion, and geospatial classification tasks.

Operates on linear optics with performance comparable to traditional neural networks.

Enables scalable, energy-efficient optical information processing for real-world applications.

Abstract

Optical computing has reemerged as a promising alternative computing paradigm for providing energy-efficient information processing in the age of artificial intelligence. Among various photonic neural network platforms, diffractive optical processing systems in free space proved high-performance computing with high parallelism. Here, we report fiber-based diffractive deep neural networks by optimizing the linear coupling of the waveguide modes. Our approach demonstrated high performance in various machine learning tasks such as biomedical disease, fashion, and geospatial classification with a simple readout layer and all-optically. Operating on linear optics, our architecture performs on par with neural networks even in complex datasets where the data cannot be separated using linear operations. These results will enable efficient and scalable diffractive information processing with waveguides for real-life computing, telecommunications, and imaging applications.