Contractive kinetic Langevin samplers beyond global Lipschitz continuity
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914188685737984 |
|---|---|
| author | Lytras, Iosif Mertikopoulos, Panayotis |
| author_facet | Lytras, Iosif Mertikopoulos, Panayotis |
| contents | In this paper, we examine the problem of sampling from log-concave distributions with (possibly) superlinear gradient growth under kinetic (underdamped) Langevin algorithms. Using a carefully tailored taming scheme, we propose two novel discretizations of the kinetic Langevin SDE, and we show that they are both contractive and satisfy a log-Sobolev inequality. Building on this, we establish a series of non-asymptotic bounds in $2$-Wasserstein distance between the law reached by each algorithm and the underlying target measure. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_12031 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Contractive kinetic Langevin samplers beyond global Lipschitz continuity Lytras, Iosif Mertikopoulos, Panayotis Probability Numerical Analysis Machine Learning math.PR, math.NA In this paper, we examine the problem of sampling from log-concave distributions with (possibly) superlinear gradient growth under kinetic (underdamped) Langevin algorithms. Using a carefully tailored taming scheme, we propose two novel discretizations of the kinetic Langevin SDE, and we show that they are both contractive and satisfy a log-Sobolev inequality. Building on this, we establish a series of non-asymptotic bounds in $2$-Wasserstein distance between the law reached by each algorithm and the underlying target measure. |
| title | Contractive kinetic Langevin samplers beyond global Lipschitz continuity |
| topic | Probability Numerical Analysis Machine Learning math.PR, math.NA |
| url | https://arxiv.org/abs/2509.12031 |