Bayesian linear regression with Student-t assumptions

TL;DR

This paper introduces a Bayesian linear regression model with Student-t assumptions (BLRS) that can be inferred exactly, generalizes existing methods, and demonstrates faster convergence in predicting online news popularity.

Contribution

It proposes an exact inference method for Bayesian linear regression with Student-t assumptions, extending the EM algorithm and proving equivalence to Gaussian-based models.

Findings

BLRS can be inferred exactly.

q-EM for BLRS converges faster than EM for BLRG.

BLRS performs well in predicting online news popularity.

Abstract

As an automatic method of determining model complexity using the training data alone, Bayesian linear regression provides us a principled way to select hyperparameters. But one often needs approximation inference if distribution assumption is beyond Gaussian distribution. In this paper, we propose a Bayesian linear regression model with Student-t assumptions (BLRS), which can be inferred exactly. In this framework, both conjugate prior and expectation maximization (EM) algorithm are generalized. Meanwhile, we prove that the maximum likelihood solution is equivalent to the standard Bayesian linear regression with Gaussian assumptions (BLRG). The $q$-EM algorithm for BLRS is nearly identical to the EM algorithm for BLRG. It is showed that $q$-EM for BLRS can converge faster than EM for BLRG for the task of predicting online news popularity.