We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ML

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: Exact Langevin Dynamics with Stochastic Gradients

Abstract: Stochastic gradient Markov Chain Monte Carlo algorithms are popular samplers for approximate inference, but they are generally biased. We show that many recent versions of these methods (e.g. Chen et al. (2014)) cannot be corrected using Metropolis-Hastings rejection sampling, because their acceptance probability is always zero. We can fix this by employing a sampler with realizable backwards trajectories, such as Gradient-Guided Monte Carlo (Horowitz, 1991), which generalizes stochastic gradient Langevin dynamics (Welling and Teh, 2011) and Hamiltonian Monte Carlo. We show that this sampler can be used with stochastic gradients, yielding nonzero acceptance probabilities, which can be computed even across multiple steps.
Comments: 13 pages, 2 figures. Accepted to the 3rd Symposium on Advances in Approximate Bayesian Inference (AABI 2021)
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG)
Cite as: arXiv:2102.01691 [stat.ML]
  (or arXiv:2102.01691v1 [stat.ML] for this version)

Submission history

From: Adrià Garriga-Alonso [view email]
[v1] Tue, 2 Feb 2021 18:59:31 GMT (185kb)

Link back to: arXiv, form interface, contact.