AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Meta-Learning MCMC Proposals

Tongzhou Wang, YI WU, Dave Moore, Stuart Russell

NeurIPS 2018 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Sequential Monte Carlo for inference · 23 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use learned block proposals instead of single-site Gibbs updatesBinary grid models · UAI 2008 inference competition grid modelsmarginal probability error over timeImproved
Use a neural block proposal instead of exact block GibbsBinary grid models · Binary grid modelserror over timeImproved
Use an approximate neural block proposal instead of exact block GibbsBinary grid models · Binary grid modelserror over epochsHurt
Use a learned block proposal instead of single-site GibbsOpen-universe GMM inference · Open-universe GMMs of various sizesmixing speedImproved
Use learned block proposals for consecutive tag variables instead of single-site GibbsNER tagging · CoNLL-2003 test sentencesF1Improved
Use learned block proposals for consecutive tag variables instead of single-site GibbsNER tagging · CoNLL-2003 test sentenceslog likelihoodImproved
Mix single-site Gibbs with neural block proposalsNER tagging · CoNLL-2003 test sentencesmixing performanceImproved
Increase the proposed tag block sizeNER tagging · CoNLL-2003 test sentencestime to mixHurt
Collapse the assignment variables and apply the Metropolis-Hastings rule to the collapsed model (ablation)Neural block proposals for open-universe GMMs · GMMs with varying numbers of observationsscalabilityImproved
Resample assignments jointly with proposed mixture variables and apply the Metropolis-Hastings rule to all proposed values (ablation)Neural block proposals for open-universe GMMs · GMMs as the number of observations growsacceptance ratioHurt
Use a mixed proposal rather than single-site Gibbs aloneNER tagging · CoNLL-2003 test sentencesF1Improved
Train on one motif instantiation distribution and apply to other parameter distributionsGrid motif proposal · Random binary-valued grid instantiationsconditional approximation divergenceImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy