☆ 4.6 Article

Scale-, Shift-, and Rotation-Invariant Diffractive Optical Networks

ACS PHOTONICS (2021)

Journal

ACS PHOTONICS

Volume 8, Issue 1, Pages 324-334

Publisher

AMER CHEMICAL SOC

DOI: 10.1021/acsphotonics.0c01583

Keywords

optical computing; diffractive networks; deep learning; optical neural networks

Funding

Fujikura (Japan)

Ask authors/readers for more resources

Protocol

Community support

Reagent

Community support

Automated Summary New
Abstract

The recent research efforts in optical computing have shifted towards developing optical neural networks that take advantage of the processing speed and parallelism of optics/photonics in machine learning applications. Diffractive Deep Neural Networks (D(2)NNs) utilize light-matter interaction over trainable surfaces, designed using deep learning, to perform statistical inference tasks as light waves propagate through them. The new training strategy introduces input object transformations as uniformly distributed random variables to improve the network's resilience against such transformations, leading to scale-, shift-, and rotation-invariant designs that are crucial for dynamic machine vision applications.

Recent research efforts in optical computing have gravitated toward developing optical neural networks that aim to benefit from the processing speed and parallelism of optics/photonics in machine learning applications. Among these endeavors, Diffractive Deep Neural Networks (D(2)NNs) harness light-matter interaction over a series of trainable surfaces, designed using deep learning, to compute a desired statistical inference task as the light waves propagate from the input plane to the output field-of-view. Although earlier studies have demonstrated the generalization capability of diffractive optical networks to unseen data, achieving, e.g., >98% image classification accuracy for handwritten digits, these previous designs are in general sensitive to the spatial scaling, translation, and rotation of the input objects. Here, we demonstrate a new training strategy for diffractive networks that introduces input object translation, rotation, and/or scaling during the training phase as uniformly distributed random variables to build resilience in their blind inference performance against such object transformations. This training strategy successfully guides the evolution of the diffractive optical network design toward a solution that is scale-, shift-, and rotation-invariant, which is especially important and useful for dynamic machine vision applications in, e.g., autonomous cars, in vivo imaging of biomedical specimen, among others.

Authors

I am an author on this paper

Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.6

Not enough ratings

Secondary Ratings

Novelty

Significance

Scientific rigor

Rate this paper

Diffractive interconnects: all-optical permutation operation using diffractive networks

Deniz Mengu, Yifan Zhao, Anika Tabassum, Mona Jarrahi, Aydogan Ozcan

Summary: This study presents the use of diffractive optical networks to perform permutation operations with a large number of input-output connections. By using deep learning techniques, the capacity of the diffractive optical network to approximate permutation operations increases with the number of diffractive layers and trainable transmission elements in the system. The authors also addressed challenges related to physical alignment and output efficiency by designing misalignment tolerant diffractive designs. The proposed diffractive permutation networks have potential applications in security, image encryption, and data processing, and can serve as channel routing and interconnection panels in wireless networks.

NANOPHOTONICS (2023)