varibayes 0.0.1

A toolbox for performing variational Bayesian inference

Variational Bayesian Inference Toolbox

---

This module is inspired by the paper ‘Black Box Variational Inference’ by Rajesh Ranganath et al. It attempts to make nearly trivial the task of fitting a variational distribution to a user-specified log-likelihood function without derivatives. Currently it only uses a mean field variational distribution, but the main class VariationalInferenceMF is flexible enough for simple subclassing in the future. This module also contains a number of implementations of stochastic gradient descent algorithms to be used for optimization.