ICPR2012 Tutorials PM-02
Penalised likelihood methods for high-dimensional pattern analysis
In recent years penalised likelihood methods, such as the least absolute shrinkage and selection operator (lasso), the elastic net, the smoothly clipped absolute deviation (SCAD) and the adaptive lasso, have become popular and been extensively studied in the statistics and machine-learning communities. These methods were developed to achieve both model fitting and feature selection, originally for regression and most recently for classification and clustering. Therefore they are extremely attractive for high-dimensional pattern analysis, in particular when sparsity is present in a so-called ‘large-p-small-n’ context, the context that the sample size, n, is smaller than the feature dimension, p, and that traditional methods often fail. In this tutorial, we shall review some established penalised likelihood methods for regression, and discuss some state-of-the-art penalised likelihood methods for classification and clustering, with respect to their intuitions, methodologies, implementations and theoretical properties.
This tutorial will review some established penalised likelihood methods for regression, and discuss some state-of-the-art penalised likelihood methods for classification and clustering, with respect to their intuitions, methodologies, implementations and theoretical properties.
Introduction to penalised likelihood methods (PLM)
PLM for high-dimensional regression (ridge regression, lasso, elastic net, SCAD, adaptive lasso, fused lasso etc.)
(a) Intuition and methodology
(c) Theoretical property
(d) Bayesian interpretation
PLM for high-dimensional classification
PLM for high-dimensional clustering
Jing-Hao Xue received the B.Eng. degree in telecommunication and information systems in 1993 and the Dr.Eng. degree in signal and information processing in 1998, both from Tsinghua University, the M.Sc. degree in medical imaging and the M.Sc. degree in statistics, both from Katholieke Universiteit Leuven in 2004, and the degree of Ph.D. in statistics from the University of Glasgow in 2008. He has worked in the Department of Statistical Science at University College London, as a Lecturer in Statistics since 2008. His research interests include statistical modelling and learning for pattern recognition, machine learning, data mining and image processing.
Archive: Call for Tutorial Proposals