On the Learnability of Offline Model-Based Optimization: A Ranking Perspective

TL;DR

This paper redefines offline model-based optimization as a ranking problem rather than a regression task, introducing a new theoretical framework and a distribution-aware ranking method that outperform existing approaches.

Contribution

It introduces a ranking-based perspective for offline MBO, develops a unified theory connecting ranking to optimization, and proposes a distribution-aware ranking method that surpasses prior methods.

Findings

Ranking-based methods outperform regression-based methods in offline MBO.

Distribution mismatch between training data and near-optimal designs is a key error source.

Intrinsic limitations exist in offline MBO, leading to unavoidable over-optimistic extrapolation.

Abstract

Offline model-based optimization (MBO) seeks to discover high-performing designs using only a fixed dataset of past evaluations. Most existing methods rely on learning a surrogate model via regression and implicitly assume that good predictive accuracy leads to good optimization performance. In this work, we challenge this assumption and study offline MBO from a learnability perspective. We argue that offline optimization is fundamentally a problem of ranking high-quality designs rather than accurate value prediction. Specifically, we introduce an optimization-oriented risk based on ranking between near-optimal and suboptimal designs, and develop a unified theoretical framework that connects surrogate learning to final optimization. We prove the theoretical advantages of ranking over regression, and identify distributional mismatch between the training data and near-optimal designs as the dominant error. Inspired by this, we design a distribution-aware ranking method to reduce this mismatch. Empirical results across various tasks show that our approach outperforms twenty existing methods, validating our theoretical findings. Additionally, both theoretical and empirical results reveal intrinsic limitations in offline MBO, showing a regime in which no offline method can avoid over-optimistic extrapolation.