Robust Freeform Metasurface Design Based on Progressively Growing Generative Networks

Abstract

A longstanding objective of machine learning-enabled inverse design is the realization of inverse neural networks that can instantaneously output a device given a desired optical function. For complex freeform devices, generative adversarial networks (GANs) can learn from images of freeform devices, but basic GAN architectures are unable to fully capture the intricate features of topologically complex structures. We show that by coupling progressive growth of the network architecture and training set with the GAN framework, generative networks can be trained to output high-performance, robust freeform metasurface devices. A combination of convolutional and self-attention layers in the network enable the accurate capture of both short-and long-range spatial patterns within topologically complex layouts. In applying this training methodology to metagratings, the best generated devices have efficiency and robustness metrics that compare with or outperform the best devices produced by gradient-based topology optimization with comparable computational cost. This study showcases the capability of generative neural networks to capture highly intricate geometric trends in physical devices, such as robustness constraints in freeform metasurfaces, and demonstrates their potential as black box inverse design tools for complex photonic technologies.

Intelligent designs in nanophotonics: from optimization towards inverse creation

“…7 (a)) characterized its working efficiency over 80%. Subsequently, more reports investigated several types of DL models [93,94] (such as generative networks [95,96]) to generate high-performance meta-gratings. An example of GAN network is presented in Fig.…”

Improving Date Fruit Classification Using CycleGAN-Generated Dataset

“…Since the introduction of GANs in 2014 [7], a number of GAN extensions have been developed, including DCGAN (Deep Convolutional GANs) [8], PGAN (Progressively Growing GANs) [9], LAPGAN (Laplacian pyramids GANs) [10], CycleGAN [11], CGAN (Conditional GANs) [12], and StyleGAN [13]. In image data augmentation, DCGANs and CycleGAN are common GAN architectures.…”

Leveraging long short-term memory (LSTM)-based neural networks for modeling structure–property relationships of metamaterials from electromagnetic responses

“…Deep learning-based models which compute abstract representations of data by learning via multiple layers of nonlinear transformations, have been the go-to approach. These models typically require curated datasets consisting of metamaterial geometries and their EM responses for training 11 , 13 – 15 . An efficient MM design scheme encompasses two aspects: the forward problem (predicting the EM response for a given geometry) as well as the inverse problem (generating the structural parameters for a desired EM response).…”