Title: Recent developments in variable selection and classification with presence-only data
Authors: Garvesh Raskutti - University of Wisconsin-Madison (United States) [presenting]
Abstract: The problem of variable selection and classification in the context of presence-only responses is addressed. Such data naturally arises in biological applications due to the high-throughput sequencing technology used. We discuss issues of estimation, inference and debiasing, and optimization relating to this problem. In particular, the imperfect labels lead to a non-convex objective which presents both statistical and optimization issues. We address these challenges, present algorithms with statistical guarantees and validate our approach on a real data application.