Rolling Diffusion Models

TL;DR

This paper introduces Rolling Diffusion, a novel temporal diffusion approach that applies a sliding window denoising process with increasing noise for later frames, improving performance on complex dynamic data.

Contribution

The paper proposes Rolling Diffusion, a new method that better models temporal uncertainty by progressively increasing noise in frames over time, outperforming standard diffusion in complex scenarios.

Findings

Rolling Diffusion outperforms standard diffusion in complex temporal data.

It improves video prediction accuracy on Kinetics-600.

It enhances fluid dynamics forecasting in chaotic systems.

Abstract

Diffusion models have recently been increasingly applied to temporal data such as video, fluid mechanics simulations, or climate data. These methods generally treat subsequent frames equally regarding the amount of noise in the diffusion process. This paper explores Rolling Diffusion: a new approach that uses a sliding window denoising process. It ensures that the diffusion process progressively corrupts through time by assigning more noise to frames that appear later in a sequence, reflecting greater uncertainty about the future as the generation process unfolds. Empirically, we show that when the temporal dynamics are complex, Rolling Diffusion is superior to standard diffusion. In particular, this result is demonstrated in a video prediction task using the Kinetics-600 video dataset and in a chaotic fluid dynamics forecasting experiment.