Novel Hybrid Techniques based on Rigorous Optimization and AI Approaches for Inverse Design of Nanophotonic Structures

This talk is focused on new techniques for the inverse design of large-scale structures in engineering that incorporate many design parameters based on combining mathematically rigorous gradient-based approaches like topological optimization with artificial-intelligence (AI) techniques for reducing the dimensionality of the design problem by several orders of magnitude. Despite the generality of these techniques, their application to the field of photonic and electromagnetic nanostructures will be discussed. Initially, a survey of AI-based approaches for design and optimization of electromagnetic nanostructures will be presented, and their shortcomings for applying to large-scale design problems will be explained. Then, the resolution of these shortcomings through combining AI tools for dimensionality reduction with gradient-based techniques for creating a meaningful set of training data will be discussed, and it is shown how such hybrid design approaches can reduce computation requirements by several orders of magnitude.

Speaker Ali Adibi, Georgia Institute of Technology

Attend in person or click here to watch on Zoom (passcode 2009A)