Stein Variational Message Passing for Continuous Graphical Models
This provides a more efficient decentralized inference algorithm for continuous graphical models, though it appears incremental as an extension of existing SVGD techniques.
The authors tackled the problem of distributed inference in continuous graphical models by extending Stein variational gradient descent to leverage Markov dependency structures, resulting in a method that outperforms standard MCMC and particle message passing baselines in empirical tests.
We propose a novel distributed inference algorithm for continuous graphical models, by extending Stein variational gradient descent (SVGD) to leverage the Markov dependency structure of the distribution of interest. Our approach combines SVGD with a set of structured local kernel functions defined on the Markov blanket of each node, which alleviates the curse of high dimensionality and simultaneously yields a distributed algorithm for decentralized inference tasks. We justify our method with theoretical analysis and show that the use of local kernels can be viewed as a new type of localized approximation that matches the target distribution on the conditional distributions of each node over its Markov blanket. Our empirical results show that our method outperforms a variety of baselines including standard MCMC and particle message passing methods.