ISTC Seminar: Manifold and Semi-supervised Learning using Laplace Operations

IS&T Center Seminar Series

April 1, 2009
Wednesday, 3 PM, CNLS Conference Room

Speaker: Professor Mikhail Belkin, Department of Computer Science and Engineering, The Ohio State University

Host:  John Hogden, Information Sciences, CCS-3

Title:  Manifold and Semi-supervised Learning using Laplace Operations
 
Contact Ron Pistone, pistone@lanl.gov, (7-3874), if you wish to meet with Mikhail during his visit.

 
Abstract:
 
In this talk I will discuss our work on manifold learning and its applications to data representation,
semi-supervised and unsupervised learning. In particular, I will discuss the significance of the
Laplace operator on the manifold and its connections to the heat diffusions. I will show some
theoretical results on reconstructing Laplace operators and their eigenfunctions from point-cloud
data and various applications, particularly to dimensionality reduction and semi-supervised learning.