Beyond Fixed Horizons: A Theoretical Framework for Adaptive Denoising Diffusions

TL;DR

This paper presents a new class of adaptive, time-homogeneous diffusion models that can dynamically adjust their steps based on noise levels, improving data generation and downstream task performance.

Contribution

It introduces a theoretical framework using Doob's $h$-transform to create adaptive diffusion processes with flexible termination criteria.

Findings

Achieves time-homogeneous diffusion processes with adaptive step counts.

Effective for generating data with low intrinsic dimensions.

Enables versatile downstream tasks like conditioning and classification.

Abstract

We introduce a new class of generative diffusion models that, unlike conventional denoising diffusion models, achieve a time-homogeneous structure for both the noising and denoising processes, allowing the number of steps to adaptively adjust based on the noise level. This is accomplished by conditioning the forward process using Doob's $h$-transform, which terminates the process at a suitable sampling distribution at a random time. The model is particularly well suited for generating data with lower intrinsic dimensions, as the termination criterion simplifies to a first-hitting rule. A key feature of the model is its adaptability to the target data, enabling a variety of downstream tasks using a pre-trained unconditional generative model. These tasks include natural conditioning through appropriate initialisation of the denoising process and classification of noisy data.