Home » Charlas Profesores Invitados » Charla de Mariano Tepper : “Clustering is semidefinitely not that hard”

Charla de Mariano Tepper : “Clustering is semidefinitely not that hard”

Lunes 20 de noviembre 16:00hs, Sala de Seminarios del Instituto de Física (7o piso) – Facultad de Ingeniería, J. Herrera y Reissig 565

El Instituto de Ingeniería Eléctrica de la Facultad de Ingeniería de la Universidad de la República invita a la siguiente charla : “Clustering is semidefinitely not that hard” por Mariano Tepper

Abstract :
In recent years, semidefinite programs (SDP) have been the subject of
interesting research in the field of clustering. In many cases, these
convex programs deliver the same answers as non-convex alternatives
and come with a guarantee of optimality. In this talk, I will argue
that SDP-KM, a popular semidefinite relaxation of K-means, can learn
manifolds present in the data, something not possible with the
original K-means formulation. To build an intuitive understanding of
SDP-KM’s manifold learning capabilities, I will present a theoretical
analysis on an idealized dataset. Additionally, SDP-KM even segregates
linearly non-separable manifolds. As generic SDP solvers are slow on
large datasets, I will also discuss the suitability of efficient
algorithms to SDP-KM. These features render SDP-KM a versatile and
interesting tool for manifold learning while remaining amenable to
theoretical analysis.

Bio :
Mariano Tepper is currently a member of the neuroscience group at the
Center for Computational Biology, Flatiron Institute. His research
focuses on image processing, computer vision, pattern recognition, and
machine learning. Previously, he was a research scientist at Duke
University. Prior to working at Duke, he was postdoctoral research
associate at the University of Minnesota. Tepper holds a Ph.D. and
licentiate degree in computer science from the Universidad de Buenos
Aires in Argentina and an M.S. in applied mathematics from the École
Normale Supérieure de Cachan in France.