Laplacians, Spectra, Kernels, and All That

John Lafferty - joint work with Guy Lebanon and Imre Kondor

Abstract

The Laplace operator is perhaps the most central mathematical object in science, so it's not surprising that it arises in machine learning. I'll describe how kernels for discrete data can be built from the Laplacian on graphs and on the Riemannian manifold defined by a family of statistical models.

Back to the Main Page