Diagnosing Suboptimal Cotangent Disintegrations in Hamiltonian Monte Carlo

2016/04/03 by Michael Betancourt, Betancourt, Michael · 7 citations
Mathematics · Decision Sciences · #Markov Chains and Monte Carlo Methods #Mathematical Approximation and Integration #Probabilistic and Robust Engineering Design

paper · pdf · doi:10.48550/arxiv.1604.00695

Abstract

When properly tuned, Hamiltonian Monte Carlo scales to some of the most challenging high-dimensional problems at the frontiers of applied statistics, but when that tuning is suboptimal the performance leaves much to be desired. In this paper I show how suboptimal choices of one critical degree of freedom, the cotangent disintegration, manifest in readily observed diagnostics that facilitate the robust application of the algorithm.

Cited by

• Forecasting Fertility with Parametric Mixture Models
• A Conceptual Introduction to Hamiltonian Monte Carlo
• City limits to partisan polarization in the American public
• Past, Present and Future of Software for Bayesian Inference
• Spatial modelling of COVID-19 incident cases using Richards' curve: an application to the Italian regions
• Modeling Complex Life Systems: Bayesian Inference for Weibull Failure Times Using Adaptive MCMC
• Improving Small-Area Estimates of Public Opinion by Calibrating to Known Population Quantities

Related

• Optimizing The Integrator Step Size for Hamiltonian Monte Carlo
• The Fundamental Incompatibility of Hamiltonian Monte Carlo and Data Subsampling
• A Conceptual Introduction to Hamiltonian Monte Carlo
• Split Hamiltonian Monte Carlo revisited
• On the Dissipation of Ideal Hamiltonian Monte Carlo Sampler