Tel-Aviv University
School of Mathematical Sciences
Department of Applied Mathematics
Applied Mathematics Seminar
special seminar; please note the different time and
place
Thursday, December 23rd, 2010
Schreiber 209, 10:10
Dan Kushnir
Yale University
Anisotropic diffusion maps of
sub-manifolds with applications
Abstract:
We introduce a method for computing extendable independent components of stochastic data sets generated by nonlinear mixing. The method relies on the spectral decomposition of a particular type of kernels constructed from observed data. The independent components are computed for a surprisingly small sub-sample of the data, and can be efficiently extended to a much larger set of new observations. We demonstrate our method performance for the empirical solution of inverse problems in electro-magnetic measurements of geological formations, and in Acoustics. This method suggests a general tool for solving empirically inverse problems.