Section: New Results

Clustering and variable selection in regression

Participants : Christophe Biernacki, Julien Jacques, Loic Yengo.

The works presented in [28] address the issue of simultaneous linear regression and clustering of predictors. A new framework is proposed that both sidesteps optimization challenges and improves prediction performance. In that framework, regression coefficients are assumed to be drawn from a gaussian mixture distribution. Prediction is thus performed using the conditional distribution of the regression coefficients given the data, while clusters are easily derived from posterior distribution in groups given the data.