Constraint-Bound Agnostic Bayesian Optimization: One Model for All Thresholds
Researchers propose CBA-BO, a learning-based framework for constrained Bayesian optimization that handles varying constraint thresholds in a single model. Unlike existing methods that treat each threshold independently, CBA-BO learns a parametric mapping from thresholds to optimal solutions, enabling direct prediction for unseen configurations without repeated optimization. A one-step Bayesian optimization refinement further improves solutions. This addresses challenges in real-world industrial design where thresholds are difficult to determine in advance and engineers need to explore different feasibility-performance trade-offs.
Key facts
- CBA-BO stands for constraint-bound agnostic Bayesian optimization
- It learns a parametric constraint model mapping thresholds to optimal solutions
- Existing constrained Bayesian optimization methods treat each threshold configuration independently
- CBA-BO predicts solutions for arbitrary unseen threshold configurations without additional optimization
- A one-step Bayesian optimization refinement further improves solutions
- The method addresses expensive constrained optimization problems in real-world industry design
- Constraint thresholds are often difficult to determine in advance in industrial design
- Engineers may need to adjust constraint thresholds to explore different feasibility-performance trade-offs
Entities
—