[YPNG] YPNG, Friday 6 Oct, 11:00-1:00, in classroom at 24 Hillhouse Ave

[David Pollard]

This Friday in the YPNG seminar Anderson Zhang will talk about the
convergence of the iterative algorithm of mean field variational
inference method.


Title: Theoretical and Computational Guarantees on Mean-Field
Variational Bayes Method for Community Detection

Abstract: The mean field variational inference approach is widely used
in statistics and machine learning for approximating posterior
distributions. But there have been very few theoretical justifications
for it. To the best of our knowledge its iterative algorithm has never
been investigated for any high dimensional/complicate model. In this
talk, we study the mean field method for community detection under the
stochastic block model. We show that the iterative algorithm (Batch
Coordinate Ascent Variational Inference) has linear convergence and
converges to the minimax optimality within log n iterations. This is a
joint work with Harrison Zhou.