Conjugate-Computation Variational Inference : Converting Variational Inference in Non-Conjugate Models to Inferences in Conjugate Models

TL;DR

The paper introduces Conjugate-computation Variational Inference (CVI), a novel algorithm that combines conjugate computations and stochastic gradients to efficiently perform variational inference in models with both conjugate and non-conjugate terms.

Contribution

It develops a new algorithm that leverages conjugate structures within variational inference, improving convergence speed in mixed models.

Findings

CVI converges faster than methods ignoring conjugate structures.

The algorithm is applicable to a wide class of models.

Experimental results demonstrate improved efficiency.

Abstract

Variational inference is computationally challenging in models that contain both conjugate and non-conjugate terms. Methods specifically designed for conjugate models, even though computationally efficient, find it difficult to deal with non-conjugate terms. On the other hand, stochastic-gradient methods can handle the non-conjugate terms but they usually ignore the conjugate structure of the model which might result in slow convergence. In this paper, we propose a new algorithm called Conjugate-computation Variational Inference (CVI) which brings the best of the two worlds together -- it uses conjugate computations for the conjugate terms and employs stochastic gradients for the rest. We derive this algorithm by using a stochastic mirror-descent method in the mean-parameter space, and then expressing each gradient step as a variational inference in a conjugate model. We demonstrate our algorithm's applicability to a large class of models and establish its convergence. Our experimental results show that our method converges much faster than the methods that ignore the conjugate structure of the model.