Variationally Inferred Sampling through a Refined Bound

In this work, a framework to boost the efficiency of Bayesian inference in probabilistic models is introduced by embedding Markov chain sampler within variational posterior approximation. We call “refined approximation”. Its strengths are its ease implementation and automatic tuning parameters, leading faster mixing time through differentiation. Several strategies approximate evidence lower bound (ELBO) computation also introduced. efficient performance showcased experimentally using state-space for time-series data, encoder density estimation conditional autoencoder as deep Bayes classifier.