Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Understanding MCMC Dynamics as Flows on the Wasserstein Space

About

It is known that the Langevin dynamics used in MCMC is the gradient flow of the KL divergence on the Wasserstein space, which helps convergence analysis and inspires recent particle-based variational inference methods (ParVIs). But no more MCMC dynamics is understood in this way. In this work, by developing novel concepts, we propose a theoretical framework that recognizes a general MCMC dynamics as the fiber-gradient Hamiltonian flow on the Wasserstein space of a fiber-Riemannian Poisson manifold. The "conservation + convergence" structure of the flow gives a clear picture on the behavior of general MCMC dynamics. The framework also enables ParVI simulation of MCMC dynamics, which enriches the ParVI family with more efficient dynamics, and also adapts ParVI advantages to MCMCs. We develop two ParVI methods for a particular MCMC dynamics and demonstrate the benefits in experiments.

Chang Liu, Jingwei Zhuo, Jun Zhu• 2019

Related benchmarks

TaskDatasetResultRank
Bayesian Neural Network InferenceBoston UCI (test)
Test Log-Likelihood-2.42
8
Bayesian Neural Network InferenceConcrete UCI (test)
Test Log-Likelihood-2.95
8
Bayesian Neural Network InferenceYacht UCI (test)
Test Log-Likelihood-0.73
8
Bayesian Neural Network InferencePower Plant (UCI) (test)
Test Log-Likelihood-2.77
8
Bayesian Neural Network InferenceEnergy UCI (test)
Test Log-Likelihood-1.36
8
Bayesian Neural Network InferenceKin8nm UCI (test)
Test Log-Likelihood1.23
8
Bayesian Neural Network InferenceYearPredictionMSD UCI (test)
Test Log-Likelihood-3.6
8
Bayesian Neural Network InferenceProtein UCI (test)
Test Log-Likelihood-3.73
8
Showing 8 of 8 rows

Other info

Follow for update