Share Email Print

Proceedings Paper

Global topology optimization neural networks for metasurface design (Conference Presentation)
Author(s): Jonathan A. Fan

Paper Abstract

I will introduce a new method for designing ultra-high efficiency metamaterials using global topology optimization networks (GLOnets). These networks combine deep generative neural networks with adjoint-based topology optimization to perform a global search and topology optimization within the design space. Importantly, these concepts utilize a population-based approach to optimize a distribution of device instances, which ensures that the full design space is properly sampled and vetted during network training. These hybrid algorithms that combine machine learning with physical calculations will set the stage for big data approaches to assist in defining the next generation of nano-based optical devices.

Paper Details

Date Published: 10 March 2020
Proc. SPIE 11290, High Contrast Metastructures IX, 112900M (10 March 2020);
Show Author Affiliations
Jonathan A. Fan, Stanford Univ. (United States)

Published in SPIE Proceedings Vol. 11290:
High Contrast Metastructures IX
Connie J. Chang-Hasnain; Andrei Faraon; Weimin Zhou, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?