Recent advances in convex clustering confer many advantages over traditional clustering methods but become limited in the face of high-dimensional data. To address this challenge, Dr. Xu has proposed biconvex clustering, a modification of convex clustering that introduces feature weights to be optimized jointly with the centroids. The method performs feature selection interpretably throughout the clustering task. He will show both theoretically and empirically that it successfully addresses the challenges of high-dimensionality in a broad range of settings.
Speaker: Jason Xu, PhD, Assistant Professor of Statistical Science, Duke University
REGISTER LINK: http://eepurl.com/g1X35P