Structured Stochastic Variational Inference
This addresses a bottleneck in scalable Bayesian inference for large datasets, offering an incremental improvement over existing methods.
The paper tackles the limitation of mean-field approximations in stochastic variational inference, which restrict posterior fidelity and introduce local optima, by relaxing these approximations to allow arbitrary dependencies, resulting in better parameter estimates with reduced bias and sensitivity.
Stochastic variational inference makes it possible to approximate posterior distributions induced by large datasets quickly using stochastic optimization. The algorithm relies on the use of fully factorized variational distributions. However, this "mean-field" independence approximation limits the fidelity of the posterior approximation, and introduces local optima. We show how to relax the mean-field approximation to allow arbitrary dependencies between global parameters and local hidden variables, producing better parameter estimates by reducing bias, sensitivity to local optima, and sensitivity to hyperparameters.