Wasserstein proximal operators describe score-based generative models and resolve memorization

[PDF] arxiv.org

Diffusion models learn low-dimensional distributions via subspace clustering

P Wang, H Zhang, Z Zhang, S Chen, Y Ma… - arXiv preprint arXiv …, 2024 - arxiv.org
Recent empirical studies have demonstrated that diffusion models can effectively learn the
image distribution and generate new samples. Remarkably, these models can achieve this …
被引用次数：18 相关文章 所有 4 个版本
[PDF] hw.ac.uk

Lipschitz-Regularized Gradient Flows and Generative Particle Algorithms for High-Dimensional Scarce Data

H Gu, P Birmpa, Y Pantazis, L Rey-Bellet… - SIAM Journal on …, 2024 - SIAM
We have developed a new class of generative algorithms capable of efficiently learning
arbitrary target distributions from possibly scarce, high-dimensional data and subsequently …
被引用次数：1 相关文章 所有 3 个版本
[PDF] arxiv.org

HJ-sampler: A Bayesian sampler for inverse problems of a stochastic process by leveraging Hamilton-Jacobi PDEs and score-based generative models

T Meng, Z Zou, J Darbon, GE Karniadakis - arXiv preprint arXiv …, 2024 - arxiv.org
The interplay between stochastic processes and optimal control has been extensively
explored in the literature. With the recent surge in the use of diffusion models, stochastic …
被引用次数：1 相关文章 所有 5 个版本
[PDF] arxiv.org

Shallow Diffuse: Robust and Invisible Watermarking through Low-Dimensional Subspaces in Diffusion Models

W Li, H Zhang, Q Qu - arXiv preprint arXiv:2410.21088, 2024 - arxiv.org
The widespread use of AI-generated content from diffusion models has raised significant
concerns regarding misinformation and copyright infringement. Watermarking is a crucial …
被引用次数：1 相关文章 所有 3 个版本
[PDF] arxiv.org

Combining Wasserstein-1 and Wasserstein-2 proximals: robust manifold learning via well-posed generative flows

H Gu, MA Katsoulakis, L Rey-Bellet… - arXiv preprint arXiv …, 2024 - arxiv.org
We formulate well-posed continuous-time generative flows for learning distributions that are
supported on low-dimensional manifolds through Wasserstein proximal regularizations of $ f …
被引用次数：1 相关文章 所有 3 个版本
[PDF] arxiv.org

Memorization and Regularization in Generative Diffusion Models

R Baptista, A Dasgupta, NB Kovachki, A Oberai… - arXiv preprint arXiv …, 2025 - arxiv.org
Diffusion models have emerged as a powerful framework for generative modeling. At the
heart of the methodology is score matching: learning gradients of families of log-densities for …
相关文章
[PDF] arxiv.org

Equivariant score-based generative models provably learn distributions with symmetries efficiently

Z Chen, MA Katsoulakis, BJ Zhang - arXiv preprint arXiv:2410.01244, 2024 - arxiv.org
Symmetry is ubiquitous in many real-world phenomena and tasks, such as physics, images,
and molecular simulations. Empirical studies have demonstrated that incorporating …
相关文章 所有 2 个版本
[PDF] arxiv.org

Nonlinear denoising score matching for enhanced learning of structured distributions

J Birrell, MA Katsoulakis, L Rey-Bellet, B Zhang… - arXiv preprint arXiv …, 2024 - arxiv.org
We present a novel method for training score-based generative models which uses
nonlinear noising dynamics to improve learning of structured distributions. Generalizing to a …
被引用次数：1 相关文章 所有 3 个版本