Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper unifies multi-objective, multi-fidelity, and constrained Bayesian optimization by framing query utility as information gain about the highest-fidelity feasible Pareto frontier. An intractable mutual information objective is approximated via a variational lower bound using mixture approximations of the Pareto-consistent region. Multi-fidelity surrogate models then enable a cost-aware acquisition function that implicitly handles fidelity selection and constraint satisfaction, validated across diverse problem settings.
A single information-theoretic framework elegantly unifies multi-objective, multi-fidelity, and constrained Bayesian optimization, bypassing separate heuristics for each.
Bayesian optimization often involves multiple objectives, constraints, and fidelity levels. We address the challenge of jointly selecting where and at which fidelity to evaluate to identify the highest-fidelity feasible Pareto frontier in this combined setting. From a unified information-theoretic perspective, we measure query utility by the information gain about this frontier, provided by an observation. Since this mutual information is intractable, we derive a variational lower bound using a mixture of under- and over-truncated approximations to the Pareto-consistent region. Multi-fidelity surrogate models propagate the information to arbitrary fidelities, yielding a cost-aware acquisition function without separate heuristics for fidelity selection or constraint handling. Experiments on synthetic, benchmark, and real-world problems demonstrate effectiveness across diverse objective, constraint, and fidelity settings.