A1817
Title: A Bayesian Lasso based sparse learning model
Authors: Ingvild Margrethe Helgoy - University of Bergen (Norway)
Yushu Li - University of Bergen (Norway) [presenting]
Abstract: The Bayesian Lasso is constructed in the linear regression framework and applies Gibbs sampling to estimate regression parameters. A new Sparse learning model, named the Bayesian Lasso Sparse (BLS) model, is developed that employs the hierarchical model formulation of the Bayesian Lasso. The main difference from the original Bayesian Lasso lies in the estimation procedure; the BLS uses a learning algorithm based on the type-II maximum likelihood procedure. In contrast to the Bayesian Lasso, the BLS provides Sparse estimates of regression parameters. The BLS is also derived for nonlinear supervised learning problems by introducing kernel functions. Comparisons of the BLS model to the Relevance Vector Machine, Fast Laplace, Bayesian Lasso, and Lasso are conducted on both simulated and real data. Numerical results demonstrate that the BLS is Sparse and precise, especially when dealing with noisy and irregular datasets.
