Manifold Learning, the Heat Equation and Spectral Clustering
Misha Belkin


We will discuss manifold learning and its connections to the heat diffusion on a manifold. In particular, we will talk about how the heat kernel yields an approximation for the Laplace-Beltrami operator, their relationship to data-dependent graph Laplacians, and algorithmic implications for unsupervised learning (dimensionality reduction and spectral clustering) and semi-supervised learning.


Slides for this lecture: PDF.

Suggested Reading