Active learning of deep surrogates for PDEs: Application to metasurface design
This addresses the problem of expensive surrogate training for photonic-device designers, offering a method to accelerate large-scale engineering optimization, though it appears incremental as it builds on existing active learning and neural-network techniques.
The paper tackled the high training cost of accurate deep surrogate models for PDEs in metasurface design by introducing an active learning algorithm that reduces required training points by over an order of magnitude, resulting in surrogate evaluations that are over two orders of magnitude faster than direct solves.
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.