Gibbs Sampling for Logistic Normal Topic Models with Graph-Based Priors

Previous work on probabilistic topic models has either focused on models with relatively simple conjugate priors that support Gibbs sampling or models with non-conjugate priors that typically require variational inference. Gibbs sampling is more accurate than variational inference and better supports the construction of composite models. We present a method for Gibbs sampling in non-conjugate logistic normal topic models, and demonstrate it on a new class of topic models with arbitrary graph-structured priors that reflect the complex relationships commonly found in document collections, while retaining simple, robust inference.