Inverse design of photonic devices with machine learning and optimization techniques

Innovative photonic devices and systems aim at achieving simultaneously a large scale of integration and high performance, leading to complex designs based on metamaterials and non-trivial geometries characterized by a large number of geometrical and material parameters. At the same time, modern cutting-edge designs usually involve multiple deterministic and stochastic figures of merit that account for both performance metrics and fabrication requirements, thus complicating the selection of the final design candidates. In this invited talk, we will discuss the potentiality of combining multi-objective analysis and optimization tools with machine learning techniques for the design of highly performing photonic devices and systems.