滚动扩散模型

Rolling Diffusion Models

February 12, 2024

作者: David Ruhe, Jonathan Heek, Tim Salimans, Emiel Hoogeboom

cs.AI

摘要

最近，扩散模型越来越多地应用于时间数据，如视频、流体力学模拟或气候数据。这些方法通常在扩散过程中平等地处理后续帧的噪声量。本文探讨了滚动扩散：一种采用滑动窗口去噪过程的新方法。它通过为序列中出现较晚的帧分配更多噪声，确保扩散过程随时间逐渐恶化，反映了随着生成过程展开，对未来存在更大不确定性。从经验上讲，我们展示了在时间动态复杂时，滚动扩散优于标准扩散。具体而言，在使用Kinetics-600视频数据集进行视频预测任务以及在混沌流体动力学预测实验中，证明了这一结果。

English

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.