Multi-scale Classification using Localized Spatial Depth

Friday, September 15, 2017 - 3:30pm

Speaker(s): 
Subhajit Dutta

Abstract: 

 In this talk, we will first discuss the notion of data depth for multivariate data and look into related inferential aspects. We shall then focus on depth based classification, and construct a classifier based on spatial depth. The construction of the proposed classifier is based on fitting a generalized additive model to the posterior probabilities corresponding to different classes. In order to cope with possible multi-modal, as well as non-elliptic nature of the population distributions, we develop a localized version of spatial depth and use that with varying degrees of localization to build a collection of classifiers. Final classification is done by aggregating over several posterior probability estimates, each of which is based on localized spatial depth with a fixed level of localization. The new classifier can be conveniently used for high-dimensional data, and its good discriminatory power for such data has been established using theoretical and some numerical results.

This is a joint work with Soham Sarkar and Anil K. Ghosh.

Seminars generally take place in 116 Old Chemistry Building on Fridays from 3:30 - 4:30 pm. For additional information contact: karen.whitesell@duke.edu or phone 919-684-8029. Sorry, but we do not have reprints available. Please feel free to contact the authors by email for follow-up information, articles, etc. 0x0AReception following seminar in 211 Old Chemistry

Old Chemistry 116

Location Info