Timezone: »

 
Poster
Analyzing Hogwild Parallel Gaussian Gibbs Sampling
Matthew Johnson · James Saunderson · Alan S Willsky

Fri Dec 06 07:00 PM -- 11:59 PM (PST) @ Harrah's Special Events Center, 2nd Floor
in Poster Session »

Sampling inference methods are computationally difficult to scale for many models in part because global dependencies can reduce opportunities for parallel computation. Without strict conditional independence structure among variables, standard Gibbs sampling theory requires sample updates to be performed sequentially, even if dependence between most variables is not strong. Empirical work has shown that some models can be sampled effectively by going "Hogwild'' and simply running Gibbs updates in parallel with only periodic global communication, but the successes and limitations of such a strategy are not well understood. As a step towards such an understanding, we study the Hogwild Gibbs sampling strategy in the context of Gaussian distributions. We develop a framework which provides convergence conditions and error bounds along with simple proofs and connections to methods in numerical linear algebra. In particular, we show that if the Gaussian precision matrix is generalized diagonally dominant, then any Hogwild Gibbs sampler, with any update schedule or allocation of variables to processors, yields a stable sampling process with the correct sample mean.

[ PDF] 

Author Information

Matthew Johnson (MIT)
James Saunderson (Massachusetts Institute of Technology)
Alan S Willsky (Massachusetts Institute of Technology)

More from the Same Authors