The Superposition of Diffusion Models Using the It\^o Density Estimator

TL;DR

This paper introduces SuperDiff, a novel framework for combining multiple pre-trained diffusion models during inference using a scalable Itô density estimator, enabling diverse and faithful image and molecule generation without retraining.

Contribution

We propose SuperDiff, a new superposition method for diffusion models that uses an Itô density estimator, allowing efficient combination of models at inference without retraining.

Findings

SuperDiff enables diverse image generation on CIFAR-10.

It improves prompt-conditioned image editing with Stable Diffusion.

SuperDiff enhances molecule and protein structure generation.

Abstract

The Cambrian explosion of easily accessible pre-trained diffusion models suggests a demand for methods that combine multiple different pre-trained diffusion models without incurring the significant computational burden of re-training a larger combined model. In this paper, we cast the problem of combining multiple pre-trained diffusion models at the generation stage under a novel proposed framework termed superposition. Theoretically, we derive superposition from rigorous first principles stemming from the celebrated continuity equation and design two novel algorithms tailor-made for combining diffusion models in SuperDiff. SuperDiff leverages a new scalable It\^o density estimator for the log likelihood of the diffusion SDE which incurs no additional overhead compared to the well-known Hutchinson's estimator needed for divergence calculations. We demonstrate that SuperDiff is scalable to large pre-trained diffusion models as superposition is performed solely through composition during inference, and also enjoys painless implementation as it combines different pre-trained vector fields through an automated re-weighting scheme. Notably, we show that SuperDiff is efficient during inference time, and mimics traditional composition operators such as the logical OR and the logical AND. We empirically demonstrate the utility of using SuperDiff for generating more diverse images on CIFAR-10, more faithful prompt conditioned image editing using Stable Diffusion, as well as improved conditional molecule generation and unconditional de novo structure design of proteins. https://github.com/necludov/super-diffusion