Reinforcement learning in a large-scale photonic recurrent neural network

Bueno, J. and Maktoobi, S. and Froehly, L. and Fischer, I. and Jacquot, M. and Larger, L. and Brunner, D. (2018) Reinforcement learning in a large-scale photonic recurrent neural network. Optica, 5 (6). pp. 756-760. ISSN 1899-7015 (https://doi.org/10.1364/OPTICA.5.000756)

[thumbnail of Bueno-etal-Optica-2018-Reinforcement-learning-in-a-large-scale-photonic-recurrent-neural-network]
Preview
 Text. Filename: Bueno_etal_Optica_2018_Reinforcement_learning_in_a_large_scale_photonic_recurrent_neural_network.pdf
Final Published Version
Download (974kB)| Preview

Abstract

Photonic neural network implementation has been gaining considerable attention as a potentially disruptive future technology. Demonstrating learning in large-scale neural networks is essential to establish photonic machine learning substrates as viable information processing systems. Realizing photonic neural networks with numerous nonlinear nodes in a fully parallel and efficient learning hardware has been lacking so far. We demonstrate a network of up to 2025 diffractively coupled photonic nodes, forming a large-scale recurrent neural network. Using a digital micro mirror device, we realize reinforcement learning. Our scheme is fully parallel, and the passive weights maximize energy efficiency and bandwidth. The computational output efficiently converges, and we achieve very good performance.

  • Item type:Article
    ID code:69578
    Dates:
    Date
    Event
    20 June 2018
    Published
    21 May 2018
    Accepted
    Keywords:neural networks, nonlinear optics, optical neural systems, Optics. Light, Electronic, Optical and Magnetic Materials, Atomic and Molecular Physics, and Optics
    Subjects:Science > Physics > Optics. Light
    Department:Faculty of Science > Physics > Institute of Photonics
    Depositing user: Pure Administrator
    Date deposited:02 Sep 2019 14:19
    Last modified:21 Mar 2023 02:35
    Related URLs:
    URI:https://strathprints.strath.ac.uk/id/eprint/69578
CORE (COnnecting REpositories)