arXiv Statistics ML
MCMC Can See Can't See
Read the original on arXiv Statistics ML →The Flow has not summarised this story yet — read it at arXiv Statistics ML.
Related stories
arXiv Statistics ML
Oct 28
Learning on-the-fly quasi-stationarity diagnostics from low-dimensional surrogates
The paper introduces a new quasi‑stationarity diagnostic for accelerated dynamics methods, replacing traditional MCMC diagnostics with a coarse‑graining framework for killed Markov jump processes and a recurrent neural network trained on particle approximations of non‑linear effective flows. It provides uniform‑in‑time, non‑asymptotic bounds for approximating the true process’s probability flow without assuming reversibility, and demonstrates that the learned RNN diagnostic outperforms the MCMC approach in ROC analysis, is less conservative, fails less often, and offers an interpretable decision threshold.
arXiv Machine Learning
Jun 15
Testing For Distribution Shifts with Conditional Conformal Test Martingales
arXiv:2602. 13848v2 Announce Type: replace Abstract: We propose a sequential test for detecting arbitrary distribution shifts that allows conformal test martingales (CTMs) to work under a fixed, reference-conditional setting.
arXiv Machine Learning
Jun 16
Asymptotically Optimal Sequential Testing with Markovian Data
arXiv:2602. 17587v3 Announce Type: replace-cross Abstract: We study one-sided and $\alpha$-correct sequential hypothesis testing for data generated by an ergodic, finite-state Markov chain.
arXiv Statistics ML
Aug 26
A Non-asymptotic Analysis for Learning and Applying a Preconditioner in MCMC
The paper presents a non‑asymptotic analysis of Markov chain Monte Carlo (MCMC) algorithms that learn and apply a preconditioner based on either the target covariance or the expected Hessian of the target potential. It compares the finite‑time computational costs of these preconditioned schemes with unpreconditioned counterparts, providing guarantees for algorithms such as the Unadjusted Langevin Algorithm (ULA) and the proximal sampler. The analysis relies on a contraction assumption in the Wasserstein‑2 distance to formalize approximate independence and bridge modern MCMC theory with classical effective sample size heuristics.
arXiv Statistics ML
Sep 17
Spectral gap of Metropolis-within-Gibbs under log-concavity
arXiv:2509. 26175v2 Announce Type: replace Abstract: The Metropolis-within-Gibbs (MwG) algorithm is a widely used Markov chain Monte Carlo method for sampling from high-dimensional distributions when exact conditional sampling is intractable.
arXiv Machine Learning
Aug 14
Dimension-Calibrated Unexplained Mass: An Interpretable Drift Statistic for Contamination Monitoring in Data Streams
arXiv:2607. 16811v4 Announce Type: replace Abstract: Drift detectors that work tend not to explain themselves, and drift detectors that explain themselves tend to fail in high dimension.