Mixed-variable policy-based optimization

TL;DR

This paper introduces a policy-based optimization method that effectively handles mixed-variable problems by combining continuous and discrete sampling, demonstrated on electromagnetics applications.

Contribution

It extends policy-based optimization to mixed-variable problems using a simple policy combination approach, enabling natural handling of both continuous and discrete variables.

Findings

Successfully solves electromagnetics problem

Achieves high-quality solutions

Demonstrates robustness in mixed-variable optimization

Abstract

The optimization of mixed-variable problems remains a significant challenge. We propose an extension of the policy-based optimization method that handles mixed-variables problems in a natural way, through a simple policy combination. This is achieved by independently sampling from a multivariate normal distribution for the continuous domain, and from multiple categorical distributions for the discrete choices. Results demonstrate that the agent successfully yields high-quality solutions on a classical problem of electromagnetics, showcasing its robustness.