Stability of fixed acquisition-function policies

Determine whether the balance between feasible-candidate discovery and feasible-hypervolume improvement is stable enough for a single fixed acquisition rule to serve throughout a constrained multi-objective Bayesian optimization campaign.

Background

Constrained multi-objective Bayesian optimization must allocate a limited evaluation budget between discovering feasible candidates and improving the hypervolume of the feasible Pareto frontier. The paper notes that fixed policies such as probability of feasibility and expected hypervolume improvement specialize in different objectives, while adaptive policies can switch between them as the campaign progresses.

The unresolved issue is whether a fixed acquisition rule can reliably maintain an appropriate balance between these competing goals, rather than requiring repeated state-dependent policy selection. The authors investigate this empirically but do not establish that the balance is stable in general.

References

However, this approach commits to a single rule for the whole campaign, which leaves open whether the balance between feasible discovery and feasible-hypervolume improvement is stable enough for a fixed rule to serve.

— Portfolio-Based Constrained Multi-Objective Bayesian Optimization for Materials Design  (2609.19550 - Sinha et al., 17 Sep 2026) in Section 1, Introduction

Agentic-Switch produced distinct state-space trajectories across the seven problems, which is at least consistent with state-conditioned selection, although our experiments cannot establish what representation supports it.

— Portfolio-Based Constrained Multi-Objective Bayesian Optimization for Materials Design  (2609.19550 - Sinha et al., 17 Sep 2026) in Section 4, Discussion