pipette
ENEnglish

MCMC Can See Can't See

Daniel Matsui Smola

PreprintCódigo disponible

En palabras de los autores

It is folklore that passing "convergence" diagnostics does not guarantee that a Markov chain has converged. We prove this as an impossibility for any finite checklist of diagnostics. Running more chains, longer chains, and smarter diagnostics does not avoid this impossibility in certifying convergence. Under regularity assumptions, we determine how the worst-case undetected non-convergence decays as more diagnostics are added, and how to choose diagnostics optimally. Folklore suggests that the sampler's slowest mixing modes are the "dangerous" ones that must be monitored; we prove that the slow modes yield nearly optimal diagnostics as sampling proceeds. Given a source class, we also give an anytime-valid bound on the remaining discrepancy and show how undetected bias caps the effective sample size that a finite checklist can certify.

Resultado principalLimitación que admiten los autores

Apareció: jueves, 15 de octubre. arXiv. Preprint, todavía sin revisión por pares.

Comentario de los autores: Code available at https://github.com/BaroqueObama/mcmc-impossibility-code