Adaptive mixtures of regressions: Improving predictive inference when population has changed

Charles Bouveyron (1) , Julien Jacques (2)

1 SAMM - Statistique, Analyse et Modélisation Multidisciplinaire (SAmos-Marin Mersenne)
2 LPP - Laboratoire Paul Painlevé - UMR 8524

The present work investigates the estimation of regression mixtures when population has changed between the training and the prediction stages. Two approaches are proposed: a parametric approach modelling the relationship between dependent variables of both populations, and a Bayesian approach in which the priors on the prediction population depend on the mixture regression parameters of the training population. The relevance of both approaches is illustrated on simulations and on an environmental dataset.

Mots clés: Transfer learning; Mixture of regressions; Switching regression; EM algorithm; Bayesian inference; MCMC algorithm; MCMC algorithm.

Domaines: Statistiques [math.ST]; Théorie [stat.TH]

Charles Bouveyron, Julien Jacques. Adaptive mixtures of regressions: Improving predictive inference when population has changed. Communications in Statistics - Simulation and Computation, 2014, 43 (10), pp.22. ⟨10.1080/03610918.2012.758737⟩. ⟨hal-00477597v3⟩