Accelerating the Design of Photonic Metamaterials by Artificial Intelligence

Abstract

In this talk, I will discuss how to accelerate the design of novel metamaterials by deep learning, a subset of artificial intelligence (AI) that learns multilevel abstraction of data using hierarchically structured layers. Different from the conventional approaches, deep learning can produce fast and accurate designs without the need of case-by-case and time-consuming numerical calculations or optimizations. The results show many exciting opportunities in the areas of optical design, integration and measurements when interfacing photonics with deep learning.