arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2504.13612cs.LGcs.AI
收藏

熵时间调度器用于生成扩散模型

Entropic Time Schedulers for Generative Diffusion Models

Radboud University(拉德堡德大学);Ghent University - imec(根特大学 - imec)

Dejan Stancevic, Florian Handke, Luca Ambrogioni

AI总结:

本文提出基于熵的时间调度器,通过重新缩放熵时间提升生成扩散模型的推理性能,无需增加函数评估次数。

摘要:

生成扩散模型的实用性能取决于噪声调度函数的适当选择,这也可以等价地表示为时间重参数化。在本文中,我们提出了一种时间调度器,该调度器基于熵选择采样点,而不是均匀时间间隔,确保每个点对最终生成贡献相等的信息。我们证明了这种时间重参数化不依赖于初始时间选择。此外,我们提供了一个可计算的精确公式,用于使用训练损失估计已训练模型的熵时间,而无需显著开销。除了熵时间外,受最优性结果的启发,我们引入了重新缩放的熵时间。在与高斯分布混合物和ImageNet的实验中,我们展示了使用(重新缩放的)熵时间显著提高了已训练模型的推理性能。特别是,我们发现,在预训练EDM2模型中,通过FID和FD-DINO评分评估的图像质量可以通过熵时间重参数化显著提高,而无需增加函数评估次数,在少量NFEs情况下改进更明显。代码可在https://github.com/DejanStancevic/Entropic-Time-Schedulers-for-Generative-Diffusion-Models上获得。

英文摘要:

The practical performance of generative diffusion models depends on the appropriate choice of the noise scheduling function, which can also be equivalently expressed as a time reparameterization. In this paper, we present a time scheduler that selects sampling points based on entropy rather than uniform time spacing, ensuring that each point contributes an equal amount of information to the final generation. We prove that this time reparameterization does not depend on the initial choice of time. Furthermore, we provide a tractable exact formula to estimate this \emph{entropic time} for a trained model using the training loss without substantial overhead. Alongside the entropic time, inspired by the optimality results, we introduce a rescaled entropic time. In our experiments with mixtures of Gaussian distributions and ImageNet, we show that using the (rescaled) entropic times greatly improves the inference performance of trained models. In particular, we found that the image quality in pretrained EDM2 models, as evaluated by FID and FD-DINO scores, can be substantially increased by the rescaled entropic time reparameterization without increasing the number of function evaluations, with greater improvements in the few NFEs regime. Code is available at https://github.com/DejanStancevic/Entropic-Time-Schedulers-for-Generative-Diffusion-Models.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑