Covariance-Adaptive Sequential Black-box Optimization for Diffusion Targeted Generation

TL;DR

This paper introduces a covariance-adaptive sequential optimization method to improve targeted generation in diffusion models using only black-box user scores, with theoretical convergence guarantees and empirical success.

Contribution

It formulates targeted diffusion model fine-tuning as a black-box optimization problem and proposes a novel algorithm with proven convergence for unknown dynamics.

Findings

Outperforms existing methods in target score achievement

Proven convergence rate of $O(rac{d^2}{ oot{2} frac{T}{}})$ for convex functions

Effective in 3D-molecule generation tasks

Abstract

Diffusion models have demonstrated great potential in generating high-quality content for images, natural language, protein domains, etc. However, how to perform user-preferred targeted generation via diffusion models with only black-box target scores of users remains challenging. To address this issue, we first formulate the fine-tuning of the targeted reserve-time stochastic differential equation (SDE) associated with a pre-trained diffusion model as a sequential black-box optimization problem. Furthermore, we propose a novel covariance-adaptive sequential optimization algorithm to optimize cumulative black-box scores under unknown transition dynamics. Theoretically, we prove a $O(\frac{d^2}{\sqrt{T}})$ convergence rate for cumulative convex functions without smooth and strongly convex assumptions. Empirically, experiments on both numerical test problems and target-guided 3D-molecule generation tasks show the superior performance of our method in achieving better target scores.