Learning Design-Score Manifold to Guide Diffusion Models for Offline Optimization

TL;DR

ManGO is a diffusion-based framework that learns the design-score manifold to enable effective offline optimization across diverse scientific and engineering domains, surpassing existing methods in accuracy and generalization.

Contribution

It introduces a novel diffusion-based approach that unifies forward prediction and backward generation for offline optimization, capturing design-score relationships holistically.

Findings

Outperforms 24 single-objective optimization methods.

Outperforms 10 multi-objective optimization methods.

Demonstrates effectiveness across various domains including materials, DNA, and engineering.

Abstract

Optimizing complex systems, from discovering therapeutic drugs to designing high-performance materials, remains a fundamental challenge across science and engineering, as the underlying rules are often unknown and costly to evaluate. Offline optimization aims to optimize designs for target scores using pre-collected datasets without system interaction. However, conventional approaches may fail beyond training data, predicting inaccurate scores and generating inferior designs. This paper introduces ManGO, a diffusion-based framework that learns the design-score manifold, capturing the design-score interdependencies holistically. Unlike existing methods that treat design and score spaces in isolation, ManGO unifies forward prediction and backward generation, attaining generalization beyond training data. Key to this is its derivative-free guidance for conditional generation, coupled with adaptive inference-time scaling that dynamically optimizes denoising paths. Extensive evaluations demonstrate that ManGO outperforms 24 single- and 10 multi-objective optimization methods across diverse domains, including synthetic tasks, robot control, material design, DNA sequence, and real-world engineering optimization.