Progressive-Growing of Generative Adversarial Networks for Metasurface Optimization
This addresses the computational cost issue in metasurface design for researchers and engineers, though it appears incremental as it builds on existing GAN methods with a new training approach.
The paper tackled the problem of generative adversarial networks (GANs) failing to capture detailed features in metasurface optimization, resulting in computationally expensive design refinement. It showed that progressively growing the network architecture and training set enables GANs to generate devices with performance comparable to gradient-based topology optimization, eliminating the need for refinement.
Generative adversarial networks, which can generate metasurfaces based on a training set of high performance device layouts, have the potential to significantly reduce the computational cost of the metasurface design process. However, basic GAN architectures are unable to fully capture the detailed features of topologically complex metasurfaces, and generated devices therefore require additional computationally-expensive design refinement. In this Letter, we show that GANs can better learn spatially fine features from high-resolution training data by progressively growing its network architecture and training set. Our results indicate that with this training methodology, the best generated devices have performances that compare well with the best devices produced by gradient-based topology optimization, thereby eliminating the need for additional design refinement. We envision that this network training method can generalize to other physical systems where device performance is strongly correlated with fine geometric structuring.