Timezone: »
Convergence of score-based generative modeling for general data distributions
Holden Lee · Jianfeng Lu · Yixin Tan
Event URL: https://openreview.net/forum?id=Sg19A8mu8sv »
We give polynomial convergence guarantees for denoising diffusion models that do not rely on the data distribution satisfying functional inequalities or strong smoothness assumptions. Assuming a $L^2$-accurate score estimate, we obtain Wasserstein distance guarantees for any distributions of bounded support or sufficiently decaying tails, as well as TV guarantees for distributions with further smoothness assumptions.
We give polynomial convergence guarantees for denoising diffusion models that do not rely on the data distribution satisfying functional inequalities or strong smoothness assumptions. Assuming a $L^2$-accurate score estimate, we obtain Wasserstein distance guarantees for any distributions of bounded support or sufficiently decaying tails, as well as TV guarantees for distributions with further smoothness assumptions.
Author Information
Holden Lee (Princeton University)
Jianfeng Lu (Duke University)
Yixin Tan (Duke University)
More from the Same Authors
-
2021 Spotlight: On the Representation of Solutions to Elliptic PDEs in Barron Spaces »
Ziang Chen · Jianfeng Lu · Yulong Lu -
2022 Panel: Panel 5B-1: Convergence for score-based… & Learning (Very) Simple… »
Sitan Chen · Yixin Tan -
2022 Poster: Convergence for score-based generative modeling with polynomial complexity »
Holden Lee · Jianfeng Lu · Yixin Tan -
2021 : Statistical Numerical PDE : Fast Rate, Neural Scaling Law and When it’s Optimal »
Yiping Lu · Haoxuan Chen · Jianfeng Lu · Lexing Ying · Jose Blanchet -
2021 Poster: On the Representation of Solutions to Elliptic PDEs in Barron Spaces »
Ziang Chen · Jianfeng Lu · Yulong Lu -
2020 Poster: A Universal Approximation Theorem of Deep Neural Networks for Expressing Probability Distributions »
Yulong Lu · Jianfeng Lu -
2019 Poster: Online sampling from log-concave distributions »
Holden Lee · Oren Mangoubi · Nisheeth Vishnoi