Active learning of deep surrogates for PDEs: Application to metasurface design

Abstract

Surrogate models for partial-differential equations are widely used in the design of meta-materials to rapidly evaluate the behavior of composable components. However, the training cost of accurate surrogates by machine learning can rapidly increase with the number of variables. For photonic-device models, we find that this training becomes especially challenging as design regions grow larger than the optical wavelength. We present an active learning algorithm that reduces the number of training points by more than an order of magnitude for a neural-network surrogate model of optical-surface components compared to random samples. Results show that the surrogate evaluation is over two orders of magnitude faster than a direct solve, and we demonstrate how this can be exploited to accelerate large-scale engineering optimization.

arXiv commentsubmitted to npj

References in corpus (20)

• Adam: A Method for Stochastic Optimization
• Dielectric Metasurfaces for Complete Control of Phase and Polarization with Subwavelength Spatial Resolution and High Transmission
• Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning
• Outlook for inverse design in nanophotonics
• Training deep neural networks for the inverse design of nanophotonic structures
• Multiwavelength Achromatic Metasurfaces by Dispersive Phase Compensation
• Freeform Diffractive Metagrating Design Based on Generative Adversarial Networks
• Inverse design of large-area metasurfaces
• Topology optimization of freeform large-area metasurfaces
• A Simple Baseline for Bayesian Uncertainty in Deep Learning
• Simulator-based training of generative models for the inverse design of metasurfaces
• Inverse designed metalenses with extended depth of focus
• Overlapping domains for topology optimization of large-area metasurfaces
• Single-Model Uncertainties for Deep Learning
• A topological encoding method for data-driven photonics inverse design
• Efficient approximation of high-dimensional functions with neural networks
• Multifunctional Metasurface Design with a Generative Adversarial Network
• Calibrating Healthcare AI: Towards Reliable and Interpretable Deep Predictive Models
• Confidence Scoring Using Whitebox Meta-models with Linear Classifier Probes
• End-to-End Nanophotonic Inverse Design for Imaging and Polarimetry

Cited by in corpus (21)

• Inverse design enables large-scale high-performance meta-optics reshaping virtual reality
• Minireview on Disordered Optical Metasurfaces
• A newcomer's guide to deep learning for inverse design in nano-photonics
• Inverse Designed Extended Depth of Focus Meta-Optics for Broadband Imaging in the Visible
• Multi-objective and categorical global optimization of photonic structures based on ResNet generative neural networks
• A neural operator-based surrogate solver for free-form electromagnetic inverse design
• Physics-enhanced deep surrogates for partial differential equations
• Deep neural networks for the prediction of the optical properties and the free-form inverse design of metamaterials
• Inverse Design of Grating Couplers Using the Policy Gradient Method from Reinforcement Learning
• Inverse Design of Multi-input Multi-output 2D Metastructured Devices
• GRIDS-Net: Inverse shape design and identification of scatterers via geometric regularization and physics-embedded deep learning
• Solving multiphysics-based inverse problems with learned surrogates and constraints
• Failure-averse Active Learning for Physics-constrained Systems
• Active Physics Informed Deep Learning: Surrogate Modeling for Non Planar Wavefront Excitation of Topological Nanophotonic Devices
• Differentiable Physics: A Position Piece
• Physics-guided hierarchical neural networks for Maxwell's equations in plasmonic metamaterials
• A Model-Constrained Tangent Slope Learning Approach for Dynamical Systems
• Efficient perturbative framework for coupling of radiative and guided modes in nearly periodic surfaces
• Fast approximate solvers for metamaterials design in electromagnetism
• Programming with Neural Surrogates of Programs
• Turaco: Complexity-Guided Data Sampling for Training Neural Surrogates of Programs