Why Label When You Can Compare? Active Constraint Pursuit in Metric Learning and Clustering

Relating pairs of samples, by way of similarity functions or
distance metrics, is at the heart of machine learning. In some
applications, such as face-photo organization in which we do not know
the identities in the photos, specifying pairwise links is the only
way to handle the learning problem"”and even naive users can provide
the annotations. In this talk, I will cover my recent work in metric
learning and active pairwise constraint pursuit. For metric learning,
I will present our efficient max-margin metric learning that learns a
Mahalanobis metric, as well as our random forest distance method that
specifies a space-varying non-linear distance function. In active
constraint pursuit for semi-supervised clustering, I will discuss how
we use the gradient of the spectral decomposition to select the next
best constraints for active user queries. I will present applications
of these methods to real data.
Corso is an associate professor of Electrical Engineering and
Computer Science at the University of Michigan. He received his PhD
and MSE degrees at The Johns Hopkins University in 2005 and 2002,
respectively, and the BS Degree with honors from Loyola College In
Maryland in 2000, all in Computer Science. He spent two years as a
post-doctoral fellow at the University of California, Los Angeles.
From 2007-14 he was a member of the Computer Science and Engineering
faculty at SUNY Buffalo. He is the recipient of the Army Research
Office Young Investigator Award 2010, NSF CAREER award 2009, SUNY
Buffalo Young Investigator Award 2011, a member of the 2009 DARPA
Computer Science Study Group, and a recipient of the Link Foundation
Fellowship in Advanced Simulation and Training 2003. Corso has
authored more than ninety peer-reviewed papers on topics of his
research interest including computer vision, robot perception, data
science, and medical imaging. He is a member of the AAAI, IEEE and
the ACM.