Oral
in
Workshop: Scaling Up Intervention Models
Amortized Active Generation of Pareto Sets
Daniel Steinberg · Asiri Wijesinghe · Rafael Oliveira · Piotr Koniusz · Cheng Soon Ong · Edwin V. Bonilla
Abstract:
We propose active generation of Pareto sets (A-GPS), a framework for online discrete black-box multi-objective optimization that learns a generative model of the Pareto set while supporting a-posteriori preference conditioning. A-GPS avoids costly hyper-volume computations and enables flexible sampling across the Pareto front without retraining. Experiments on synthetic functions and protein design tasks show strong sample efficiency and effective preference incorporation.
Chat is not available.