Amortized Active Generation of Pareto Sets

Daniel M. Steinberg, Asiri Wijesinghe, Rafael Oliveira, Piotr Koniusz, Cheng Soon Ong, Edwin V. Bonilla

Keywords: Multi-objective optimization, active generation

TL;DR: An active generation method that supports preference conditioning

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.

Submission Number: 34