First, I will briefly talk about general principles of formulating algorithms for topographic maps of vectorial data. I will then discuss possibilities of extending such topographic maps to more complex structured data types such as sequences or graphs. The emphasis will be placed on generative probabilistic modeling. In this framework, topographic maps are realized as constrained mixture of appropriate noise models. The approach will be illustrated on chorals by J.S. Bach, pictograms represented as directed acyclic graphs and fluxes from eclipsing binary stars.
Last modified: Thursday, 12-Dec-2013 09:32:02 NZDT
This page is maintained by the seminar list administrator.