Ed George is this year's Cornell Distinguished Alumni. He is currently a professor of statistics at the Wharton School of the University of Pennsylvania. His research interests include hierarchical modeling, model uncertainty, shrinkage estimation, treed modeling, variable selection, and wavelet regression.Having received his bachelor's degree in mathematics from Cornell, George went on to earn a master's from SUNY at Stony Brook and a PhD from Stanford University. He has held previous appointments at the University of Texas at Austin and the University of Chicago.
Among his many accomplishments, Dr. George is a fellow of the International Society for Bayesian Analysis (2014), the American Statistical Association (1997), and the Institute of Mathematical Statistics (1995). He also served as editor of the Annals of Statistics from 2016 to 2018, and executive editor of Statistical Science from 2004 to 2007.
Talk: Bayesian Penalty Mixing with the Spike and Slab Lasso
Abstract: Despite the wide adoption of spike-and-slab methodology for Bayesian variable selection, its potential for penalized likelihood estimation has largely been overlooked. We bridge this gap by cross-fertilizing these two paradigms with the Spike-and-Slab Lasso, a procedure for simultaneous variable selection and parameter estimation in linear regression. A mixture of two Laplace distributions, the Spike-and-Slab Lasso prior induces a new class of self-adaptive penalty functions that arise from a fully Bayes spike-and-slab formulation, ultimately moving beyond the separable penalty framework. A virtue of these non-separable penalties is their ability to borrow strength across coordinates, adapt to ensemble sparsity information and exert multiplicity adjustment. With a path-following scheme for dynamic posterior exploration, efficient EM and coordinatewise implementations, the fully Bayes penalty is seen to mimic oracle performance, providing a viable alternative to cross-validation. Further elaborations of the Spike-and-Slab Lasso for fast Bayesian factor analysis illuminate its broad potential. (This is joint work with Veronika Rockova).