Research on Bayesian Model Averaging for Lasso based on Analysis of Scientific Materials

The Lasso (least absolute shrinkage and selection operator) estimates a vector of regression coefficients by minimizing the residual sum of squares subject to a constraint on the L-1-norm of coefficient vector, which has been an attractive technique for regularization and variable selection. In this paper, we study the Bayesian Model Averaging(BMA) for Lasso, which accounts for the uncertainty about the best model to choose by averaging over multiple models. Experimental results on simulated data show that BMA has significant advantage over the model selection method based on Bayesian information criterion (BIC).