Inverse design of the transmission matrix in a random system using Reinforcement Learning

TL;DR

This paper introduces a reinforcement learning approach to inverse design of scattering systems by modifying transmission matrices, enabling control over power conversion, eigenvalue degeneracies, and mode conversion in complex systems.

Contribution

It applies reinforcement learning, specifically Proximal Policy Optimization, to the inverse design of transmission matrices in scattering systems, addressing a highly non-convex optimization landscape.

Findings

Designed transmission matrices for power conversion and zero-transmission modes.

Achieved exceptional points with degenerate eigenvalues and unidirectional mode conversion.

Enforced uniform channel participation in degenerate eigenvalue scenarios.

Abstract

This work presents an approach to the inverse design of scattering systems by modifying the transmission matrix using reinforcement learning. We utilize Proximal Policy Optimization to navigate the highly non-convex landscape of the object function to achieve three types of transmission matrices: (1) Fixed-ratio power conversion and zero-transmission mode in rank-1 matrices, (2) exceptional points with degenerate eigenvalues and unidirectional mode conversion, and (3) uniform channel participation is enforced when transmission eigenvalues are degenerate.