Mar 10, 2024 · In this paper, we investigate the prior diffusion technique for the target distributions satisfying log-Sobolev inequality (LSI), which covers a much broader ...

Mar 10, 2024 · Beyond the log-concavity, it also includes non-log-concave distributions, such as mixtures and perturbations of log-concave distributions, which ...

Mar 12, 2024 · An Improved Analysis of Langevin Algorithms with Prior Diffusion for Non-Log-Concave Sampling https://t.co/qkK3va8K9M.

Prior diffusion in Langevin algorithms enables dimension-independent convergence for non-log-concave distributions.

图表 · 解决问题. 研究高维采样问题中计算复杂度与维度依赖性的关系，尤其是针对log-Sobolev不等式的目标分布。 · 关键思路 · 其它亮点 · 相关研究.

This work proposes the symmetrized Langevin algorithm (SLA), which should have a smaller bias than ULA, at the price of implementing a proximal gradient ...

In this section, we describe our new convergence results for the proximal sampler under various assumptions, beginning with the strongly log-concave and weakly ...

In this paper, we investigate the prior diffusion technique for the target distributions satisfying log-Sobolev inequality (LSI), which covers a much broader ...

In this paper, we provide non-asymptotic upper bounds on the error of sampling from a tar- get density over Rp using three schemes of discretized Langevin ...

Sep 6, 2024 · Our proof technique provides a new way to study the convergence of Langevin based algorithms, and sheds some light on the design of fast ...