Mathematical treatment of large dimensionality data

Sultan Valeev

Department of Applied Mathematics and Informatics, Ulyanovsk State Technical University, Sev. Venets str., 32 Ylyanovsk, 432057, Russia, e-mail: sgv@ulstu.ru

In a fact of increasing precision of observations the MLS’s problems – evaluation of parameters - is solving for models with enlarging “enormous” dimensionality. There are at least 3 problems: - insufficiency of hardware capabilities; - incorrectness of postulated processing model, which can include up to 90% of uninformative and duplicated components. It requires full exhaustion of 2^p competitive structures by specified quality criterion for reducing dimensionality, where p is a quantity of used parameters; -violation of Gauss-Markov conditions while using MLS. Methods for solving these problems with mathematical, software and hardware ways are considering.