GLISp-r: A preference-based optimization algorithm with convergence guarantees

Preference-based optimization algorithms are iterative procedures that seek the optimal value for a decision variable based only on comparisons between couples of different samples. At each iteration, a human decision-maker is asked to express a preference between two samples, highlighting which one, if any, is better than the other. The optimization procedure must use the observed preferences to find the value of the decision variable that is most preferred by the human decision-maker, while also minimizing the number of comparisons. In this work, we propose GLISp-r, an extension of a recent preference-based optimization procedure called GLISp. The latter uses a Radial Basis Function surrogate to describe the tastes of the individual. Iteratively, GLISp proposes new samples to compare with the current best candidate by trading off exploitation of the surrogate model and exploration of the decision space. In GLISp-r, we propose a different criterion to use when looking for a new candidate sample that is inspired by MSRS, a popular procedure in the black-box optimization framework (which is closely related to the preference-based one). Compared to GLISp, GLISp-r is less likely to get stuck on local optimizers of the preference-based optimization problem. We motivate this claim theoretically, with a proof of convergence, and empirically, by comparing the performances of GLISp and GLISp-r on different benchmark optimization problems.