Doubly Stochastic Variational Bayes for non-Conjugate Inference
Michalis K. Titsias, Miguel L zaro-gredilla · 2014
We propose a simple and effective variational inference algorithm based on stochastic optimi-sation that can be widely applied for Bayesian non-conjugate inference in continuous parameter spaces. This algorithm is based on stochastic ap-proximation and allows for efficient use of gra-dient information from the model joint density. We demonstrate these properties using illustra-tive examples as well as in challenging and di-verse Bayesian inference problems such as vari-able selection in logistic regression and fully Bayesian inference over kernel hyperparameters in Gaussian process regression. 1.