What question did this study set out to answer?

The aim is to introduce optorch, a framework for simulating and deploying diffractive deep neural networks (D2NNs).

May 24, 20260 citationsOpen Access

optorch: An Open-Source PyTorch Framework for Diffractive Deep Neural Networks

Key Points

The aim is to introduce optorch, a framework for simulating and deploying diffractive deep neural networks (D2NNs).
Developed an open-source Python framework using PyTorch for D2NN simulation and training.
Utilized the Angular Spectrum Method for simulating light propagation.
Trained a 3-layer D2NN on the MNIST dataset achieving fast training times on standard CPUs.
Achieved 89.79% test accuracy on the MNIST dataset with only 2,352 learnable parameters.
Training time was approximately 2 minutes on a standard CPU.
Performance within 2% of the 2018 MIT Science paper while using 340 times fewer parameters.

Abstract

optorch is an open-source Python framework for simulating, training, and deploying Diffractive Deep Neural Networks (D2NNs). Unlike conventional neural networks that rely on electronic matrix multiplication, optical neural networks compute through the physical diffraction of light, enabling ultra-low-latency and energy-efficient AI inference. Built on the Angular Spectrum Method, optorch provides fully differentiable light propagation simulation in a PyTorch-native workflow. A 3-layer D2NN trained with optorch achieves 89.79% MNIST test accuracy using only 2,352 learnable parameters, training in approximately 2 minutes on a standard CPU — within 2% of the original 2018 MIT Science paper while using 340× fewer parameters. optorch is the first pip-installable, PyTorch-native open-source framework for optical neural networks. GitHub: optorch on GitHub

问 AI

Bookmark

View Full Paper