eduzhai > Applied Sciences > Engineering >

Automated Optical Multi-layer Design via Deep Reinforcement Learning

  • Save

... pages left unread,continue reading

Document pages: 15 pages

Abstract: Optical multi-layer thin films are widely used in optical and energyapplications requiring photonic designs. Engineers often design such structuresbased on their physical intuition. However, solely relying on human experts canbe time-consuming and may lead to sub-optimal designs, especially when thedesign space is large. In this work, we frame the multi-layer optical designtask as a sequence generation problem. A deep sequence generation network isproposed for efficiently generating optical layer sequences. We train the deepsequence generation network with proximal policy optimization to generatemulti-layer structures with desired properties. The proposed method is appliedto two energy applications. Our algorithm successfully discoveredhigh-performance designs, outperforming structures designed by human experts intask 1, and a state-of-the-art memetic algorithm in task 2.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×