Plackett-Luce regression: A new Bayesian model for polychotomous data

TL;DR

This paper introduces a Bayesian Plackett-Luce regression model for polychotomous data, offering a simpler, efficient alternative to traditional multinomial logistic regression with sparse feature learning.

Contribution

It develops a novel Bayesian model based on the Plackett-Luce framework, with EM, Gibbs sampling, and variational inference methods, improving simplicity and sparsity in polychotomous data analysis.

Findings

Competitive performance with sparse Bayesian multinomial logistic regression

Effective in modeling polychotomous data

Simpler implementation with standard distributions

Abstract

Multinomial logistic regression is one of the most popular models for modelling the effect of explanatory variables on a subject choice between a set of specified options. This model has found numerous applications in machine learning, psychology or economy. Bayesian inference in this model is non trivial and requires, either to resort to a MetropolisHastings algorithm, or rejection sampling within a Gibbs sampler. In this paper, we propose an alternative model to multinomial logistic regression. The model builds on the Plackett-Luce model, a popular model for multiple comparisons. We show that the introduction of a suitable set of auxiliary variables leads to an Expectation-Maximization algorithm to find Maximum A Posteriori estimates of the parameters. We further provide a full Bayesian treatment by deriving a Gibbs sampler, which only requires to sample from highly standard distributions. We also propose a variational approximate inference scheme. All are very simple to implement. One property of our Plackett-Luce regression model is that it learns a sparse set of feature weights. We compare our method to sparse Bayesian multinomial logistic regression and show that it is competitive, especially in presence of polychotomous data.