Covariance-Adaptive Sequential Black-box Optimization for Diffusion Targeted Generation
This addresses the problem of targeted content generation for users in domains like images and molecules, but it is incremental as it builds on existing diffusion model frameworks.
The paper tackles the challenge of performing user-preferred targeted generation with diffusion models using only black-box target scores, by formulating it as a sequential black-box optimization problem and proposing a covariance-adaptive algorithm. The result includes a theoretical convergence rate of O(d^2/√T) and empirical superiority in target-guided 3D-molecule generation tasks.
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.