Mathematical Physics Seminar

We give a brief survey of some of our recent work on
approximation of functions of unstructured, high dimensional data sets. Ideas from diffusion geometry have been used widely in this context. Typically, they involve the computation of eigenvalues and eigenvectors of a large matrix. We will discuss our efforts to avoid or simplify these
computations. We will demonstrate some applications, and if time permits, discuss a new theory regarding wavelet–like representations on graphs and trees.