DisDiff: Unsupervised Disentanglement of Diffusion Probabilistic Models

TL;DR

DisDiff introduces an unsupervised method to automatically discover and disentangle underlying factors in diffusion probabilistic models, enhancing interpretability and controllability of generative processes without requiring annotations.

Contribution

The paper proposes DisDiff, the first unsupervised approach to disentangle factors in pre-trained DPMs by decomposing gradient fields, enabling explicit factor representation.

Findings

DisDiff successfully disentangles factors in synthetic datasets.

DisDiff effectively discovers inherent factors in real-world data.

The method improves interpretability of diffusion models.

Abstract

Targeting to understand the underlying explainable factors behind observations and modeling the conditional generation process on these factors, we connect disentangled representation learning to Diffusion Probabilistic Models (DPMs) to take advantage of the remarkable modeling ability of DPMs. We propose a new task, disentanglement of (DPMs): given a pre-trained DPM, without any annotations of the factors, the task is to automatically discover the inherent factors behind the observations and disentangle the gradient fields of DPM into sub-gradient fields, each conditioned on the representation of each discovered factor. With disentangled DPMs, those inherent factors can be automatically discovered, explicitly represented, and clearly injected into the diffusion process via the sub-gradient fields. To tackle this task, we devise an unsupervised approach named DisDiff, achieving disentangled representation learning in the framework of DPMs. Extensive experiments on synthetic and real-world datasets demonstrate the effectiveness of DisDiff.