Department of Statistics, University of Warwick, UK
Wednesday 23 May 2007, 16:00
Seminar Room B10 (Basement)
Alexandra House, 17 Queen Square, London, WC1N 3AR
Bayesian adaptive lassos with non-convex penalization
The lasso has sparked interest in the use of penalization of the log-likelihood for feature selection. Recently, there have been attempts to propose penalty functions which improve upon the Lassos properties for variable selection and prediction, such as SCAD and the Adaptive Lasso. We adopt the Bayesian interpretation of the Lasso as the maximum a posteriori (MAP) and explore changing the prior distribution. We are particularly interested in the more variable than observations case. Our methodology can give rise to multiple modes of the posterior distribution.
A paper is available at