Prestigious journal Optica features on its cover an article with participation of IFISC
Researchers of
FEMTO-ST and IFISC demonstrate a large-scale photonic recurrent neural network
with reinforcement learning and publish results in
prestigious journal Optica, featured on its cover page.
Researchers of FEMTO-ST and IFISC demonstrate a large-scale photonic recurrent neural network with reinforcement learning and publish results in prestigious journal Optica, featured on its cover page.
June 25, 2018
The article "Reinforcement learning in a large-scale photonic recurrent neural network” by J. Bueno, S. Maktoobi, L. Froehly, I. Fischer, M. Jacquot, L. Larger, and D. Brunner has been selected by the journal Optica, edited by OSA (Optical Society of America), as the cover of its volume 5, issue 6.
Photonic neural network implementations have been gaining considerable attention as a potentially disruptive future technology. However, photonic neural networks with many nonlinear nodes in a fully parallel and efficient learning hardware have been lacking so far. In their new publication, researchers at FEMTO-ST (Besançon, France) and IFISC (Palma, Spain) demonstrate a network of up to 2025 optically coupled nodes, forming a large-scale recurrent neural network. Implementing reinforcement learning, they obtain very good performance in a nonlinear prediction task. Ultimately, their work opens the door to novel and versatile photonic neural networks concepts.
The research is illustrated on the cover of the journal with the image of a “digital micro mirror device trained by reinforcement learning for photonic neural network implementation”.
