Statistics > Machine Learning
[Submitted on 29 Aug 2016 (this version), latest version 9 May 2021 (v4)]
Title:Manifold Langevin Monte Carlo with Partial Metric Updates
View PDFAbstract:Manifold Markov chain Monte Carlo algorithms have been introduced to sample more effectively from challenging target densities exhibiting multiple modes or strong correlations. Such algorithms exploit the local geometry of the parameter space, thus enabling chains to achieve a faster convergence rate when measured in number of steps. However, often acquiring local geometric information increases computational complexity per step to the extent that sampling from high-dimensional targets becomes inefficient in terms of total computational time. This paper proposes a manifold Langevin Monte Carlo framework aimed at balancing the benefits of exploiting local geometry with computational requirements to achieve a high effective sample size for a given computational cost. The suggested strategy regulates the frequency of manifold-based updates via a schedule. An exponentially decaying schedule is put forward that enables more frequent updates of geometric information in early transient phases of the chain, while saving computational time in late stationary phases. Alternatively, a modulo schedule is introduced to allow for infrequent yet recurring geometric updates throughout the sampling course, acting to adaptively update proposals to learn the covariance structure of the parameter space. The average complexity can be manually set for either of these two schedules depending on the need for geometric exploitation posed by the underlying model.
Submission history
From: Theodore Papamarkou [view email][v1] Mon, 29 Aug 2016 10:32:15 UTC (5,830 KB)
[v2] Sun, 4 Jun 2017 18:56:51 UTC (807 KB)
[v3] Sat, 5 May 2018 20:21:14 UTC (3,764 KB)
[v4] Sun, 9 May 2021 05:50:54 UTC (3,736 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
Connected Papers (What is Connected Papers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.