arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2209.12381cs.LGmath.PRmath.STstat.MLstat.TH
收藏

一般数据分布基于得分的生成式建模的收敛性

Convergence of score-based generative modeling for general data distributions

Johns Hopkins University(约翰斯·霍普金斯大学);Duke University(杜克大学)

Holden Lee, Jianfeng Lu, Yixin Tan

摘要:

基于分数的生成建模(SGM)已经发展成为一种非常成功的方法,用于学习从复杂的数据分布(如图像和音频)中生成样本。它是基于进化一个将白噪声转换为来自学习分布的样本,使用评分函数或梯度log-pdf的估计值。以前的收敛性分析,这些方法都遭受了强假设的数据分布或指数依赖关系,因此无法提供有效的保证,在实践中出现的多峰和非光滑分布,并观察到良好的经验性能。我们考虑一种流行的SGM -去噪扩散模型-并给出一般数据分布的多项式收敛保证,没有与函数不等式或光滑性相关的假设。假设$L^2$ -准确的分数估计,我们得到Wasserstein距离保证任何分布的有界支持或充分衰减的尾巴,以及电视保证分布进一步平滑的假设。

英文摘要:

Score-based generative modeling (SGM) has grown to be a hugely successful method for learning to generate samples from complex data distributions such as that of images and audio. It is based on evolving an SDE that transforms white noise into a sample from the learned distribution, using estimates of the score function, or gradient log-pdf. Previous convergence analyses for these methods have suffered either from strong assumptions on the data distribution or exponential dependencies, and hence fail to give efficient guarantees for the multimodal and non-smooth distributions that arise in practice and for which good empirical performance is observed. We consider a popular kind of SGM -- denoising diffusion models -- and give polynomial convergence guarantees for general data distributions, with no assumptions related to functional inequalities or smoothness. Assuming $L^2$-accurate score estimates, we obtain Wasserstein distance guarantees for any distribution of bounded support or sufficiently decaying tails, as well as TV guarantees for distributions with further smoothness assumptions.

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。

↑