Research map / Optimization / Nonconvex and smooth optimization
Exponential ergodicity of mirror-Langevin diffusions
Sinho Chewi, Thibaut Le Gouic, Chen Lu, Tyler Maunu, Philippe Rigollet, Austin Stromme
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace Langevin diffusion with Newton-Langevin algorithm (mirror map = potential) | Newton-Langevin Algorithm vs ULA/TULA · Generalized Gaussian (gamma=3/4), d=100, Sigma=diag(1..100) | squared mean error and relative scatter error | Improved |
| use a large step size h=0.7 for all samplers (ablation) | Newton-Langevin Algorithm vs ULA and TULA · ill-conditioned Gaussian d=100 | sample quality (qualitative) | Improved |
| reduce step size to 0.05 for all samplers (ablation) | ULA and TULA · ill-conditioned Gaussian d=100 | sample quality (qualitative) | Mixed |
| replace ULA/TULA with Newton-Langevin | Newton-Langevin Algorithm · 2D Bayesian logistic regression posterior | sample quality (qualitative) | Improved |
| decrease step size from 0.05 to 0.01 in logistic regression (ablation) | ULA and TULA · Bayesian logistic regression | exploration of posterior (qualitative): 0.05 -> 0.01 | No clear effect |
| sample a thin rectangle via Newton-Langevin on a barrier-approximated target instead of projected Langevin | Newton-Langevin vs PLA and MALA · uniform on [-0.01,0.01]x[-1,1] | W2 distance | Improved |
| use larger step size for PLA (ablation) | Projected Langevin Algorithm · thin rectangle uniform sampling | sample distribution | Hurt |
Closest papers by meaning
- Langevin Monte Carlo for strongly log-concave distributions: Randomized midpoint revisited (ICLR 2024 · similarity 0.88)
- Wasserstein Convergence of Critically Damped Langevin Diffusions (NeurIPS 2025 · similarity 0.87)
- The Poisson Midpoint Method for Langevin Dynamics: Provably Efficient Discretization for Diffusion Models (NeurIPS 2024 · similarity 0.87)
- Scalable Discrete Diffusion Samplers: Combinatorial Optimization and Statistical Physics (ICLR 2025 · similarity 0.86)
- Provable Convergence and Limitations of Geometric Tempering for Langevin Dynamics (ICLR 2025 · similarity 0.86)
- Sequential Controlled Langevin Diffusions (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
