Home > Events > 2013 Seminars & Colloquia > BOAZ NADLER - Weizmann Institute

BOAZ NADLER - Weizmann Institute

Main Content

An introduction to Diffusion Maps, Applications to Spectral Clustering, Manifold Learning, and more
14 March 2013
327 Thomas Bldg.
Add event to calendar

In many scientific fields, there is a need to analyze high-dimensional data. One common assumption is that despite being
high-dimensional, the data have some "simple" underlying structure. If the data are assumed to reside in a low-dimensional linear subspace then this leads to the popular principal component analysis (PCA) method.

In this talk, we will present an introduction to diffusion maps - a framework for analyzing data that have a low-dimensional, but possibly non-linear, intrinsic structure (be it manifold, or a union of clusters), or even non-Euclidean data that are encoded as a graph.

We will present relations and connections to several popular algorithms in machine learning as well as some limitations of this global approach.

Filed under: ,