Active Preference Learning over Latent Preference Archetypes for Many-Objective Bayesian Optimization

The paper introduces an active preference learning framework for many-objective Bayesian optimization that models preferences as a Dirichlet-process mixture of latent archetypes. It uses mixture-aware information-theoretic query strategies to separately identify archetypes and refine preferences within each archetype, employing a hybrid acquisition policy. Experiments on synthetic benchmarks and a real-world chemical process design case study show that this approach outperforms existing preference-based Bayesian optimization methods and recovers interpretable latent preference structures.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Machine Learning.

• benchmarks