Title: Scalable and Model-free Methods for Multiclass Probability Estimation
Abstract:
Classical approaches for multiclass probability estimation are mostly model-based, such as logistic regression or LDA, by making certain assumptions on the underlying data distribution. We propose a new class of model-free methods to estimate class probabilities based on large-margin classifiers. The method is scalable for high-dimensional data by employing the divide-and-conquer technique, which solves multiple weighted large-margin classifiers and then constructs probability estimates by aggregating multiple classification rules. Without relying on any parametric assumption, the estimates are shown to be consistent asymptotically. Both simulated and real data examples are presented to illustrate the performance of the new procedure.
Bio:
Hao Helen Zhang is a Professor of Department of Mathematics at University of Arizona, as well as a faculty member of Statistics Graduate Interdisciplinary Program (GIDP). Dr. Zhang obtained a Ph.D. in Statistics from University of Wisconsin at Madison in 2002. She was assistant and associate professor of Statistics at North Carolina State University 2002-2011. Dr. Zhang’s research areas include statistical machine learning, high-dimensional data analysis, nonparametric smoothing, and biomedical data analysis. Her research was funded by NSF, NIH, NSA, including a NSF CAREER Award. Dr. Zhang is currently Associate Editor of Journal of American Statistical Association, Journal of Computational and Graphical Statistics, and Statistical Analysis and Data Mining. She is a Fellow of the American Statistical Association, Fellow of the Institute of Mathematical Statistics, as well as elected member of the International Statistical Institute.
