Non-asymptotic guarantees for sampling by stochastic gradient descent

DOI10.3103/S1068362319020031zbMath1436.62044arXiv1811.00781OpenAlexW2964311113MaRDI QIDQ2290072

Avetik Karagulyan

Publication date: 27 January 2020

Published in: Journal of Contemporary Mathematical Analysis. Armenian Academy of Sciences (Search for Journal in Brave)

Full work available at URL: https://arxiv.org/abs/1811.00781

zbMATH Keywords

Markov chain Monte Carloconvergence rategradient descentLangevin algorithmapproximate sampling

Mathematics Subject Classification ID

Sampling theory, sample surveys (62D05) Monte Carlo methods (65C05) Asymptotic approximations, asymptotic expansions (steepest descent, etc.) (41A60)

Cites Work

• Exponential convergence of Langevin distributions and their discrete approximations
• Langevin diffusions and Metropolis-Hastings algorithms
• Geometric ergodicity of Metropolis algorithms
• Optimal scaling and diffusion limits for the Langevin algorithm in high dimensions
• Langevin-type models. I: Diffusions with given stationary distributions and their discretizations
• High-dimensional Bayesian inference via the unadjusted Langevin algorithm
• Nonasymptotic convergence analysis for the unadjusted Langevin algorithm
• Rate of convergence for ergodic continuous Markov processes: Lyapunov versus Poincaré
• Recursive Stochastic Algorithms for Global Optimization in $\mathbb{R}^d $
• Optimal Scaling of Discrete Approximations to Langevin Diffusions
• Optimization Methods for Large-Scale Machine Learning
• Theoretical Guarantees for Approximate Sampling from Smooth and Log-Concave Densities