Pith. sign in

REVIEW 3 cited by

Convergence Analysis of Stochastic Gradient Descent with MCMC Estimators

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2303.10599 v2 pith:5WVWP35D submitted 2023-03-19 stat.ML math.OC

Convergence Analysis of Stochastic Gradient Descent with MCMC Estimators

classification stat.ML math.OC
keywords mcmcepsilonconvergenceestimatorgradientmcmc-sgdpointsstochastic
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Understanding stochastic gradient descent (SGD) and its variants is essential for machine learning. However, most of the preceding analyses are conducted under amenable conditions such as unbiased gradient estimator and bounded objective functions, which does not encompass many sophisticated applications, such as variational Monte Carlo, entropy-regularized reinforcement learning and variational inference. In this paper, we consider the SGD algorithm that employ the Markov Chain Monte Carlo (MCMC) estimator to compute the gradient, called MCMC-SGD. Since MCMC reduces the sampling complexity significantly, it is an asymptotically convergent biased estimator in practice. Moreover, by incorporating a general class of unbounded functions, it is much more difficult to analyze the MCMC sampling error. Therefore, we assume that the function is sub-exponential and use the Bernstein inequality for non-stationary Markov chains to derive error bounds of the MCMC estimator. Consequently, MCMC-SGD is proven to have a first order convergence rate $O(\log K/\sqrt{n K})$ with $K$ iterations and a sample size $n$. It partially explains how MCMC influences the behavior of SGD. Furthermore, we verify the correlated negative curvature condition under reasonable assumptions. It is shown that MCMC-SGD escapes from saddle points and reaches $(\epsilon,\epsilon^{1/4})$ approximate second order stationary points or $\epsilon^{1/2}$-variance points at least $O(\epsilon^{-11/2}\log^{2}(1/\epsilon) )$ steps with high probability. Our analysis unveils the convergence pattern of MCMC-SGD across a broad class of stochastic optimization problems, and interprets the convergence phenomena observed in practical applications.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Is Variational Monte Carlo Robust? Sharp Moment Thresholds and Heavy-tailed Stochastic Optimization

    cs.LG 2026-06 unverdicted novelty 7.0

    VMC local energy and gradient estimators are generically heavy-tailed for common ansatze due to nodal sets, but a new clipped variant converges in the low-moment regime.

  2. Is Variational Monte Carlo Robust? Sharp Moment Thresholds and Heavy-tailed Stochastic Optimization

    cs.LG 2026-06 conditional novelty 7.0

    VMC's gradient estimators are generically heavy-tailed (no 3/2 moment for Slater–Jastrow); PS-Clip-VMC, which clips energies and per-sample gradients, is provably convergent under weak moments and stabilizes FermiNet ...

  3. Momentum Stability and Adaptive Control in Stochastic Reconfiguration

    math.OC 2026-04 unverdicted novelty 5.0

    Convergence holds for momentum μ less than 1 in SPRING under mild assumptions, but μ=1 risks divergence; PRIME-SR adapts momentum via spectral dimension and subspace overlap to match tuned performance with better robustness.