Factorization machine with quantum annealing (FMQA) is a widely used python package for solving black‐box optimization problems using D‐Wave quantum annealers. It has been successfully applied, e.g., to metamaterial design, polymer design, and wing shape optimization. Batch black‐box optimization, where an objective function is evaluated at multiple points simultaneously, appears commonly in scientific design problems. FMQA's batch optimization option relies on sampling via quantum noise, which is not controllable or effective. This paper presents a new python package, factorization machine with iterative quantum reverse annealing (FMIRA). Multiple particles in the design space are prepared, and each of them is independently updated via reverse annealing with respect to a local Ising model obtained from a FM. In the batch settings, FMIRA showed better performance in multiple benchmark problems.
Tučs et al. (Thu,) studied this question.