This morning we attended Professor Shary’s seminar entitled:
Data fitting under interval imprecision and uncertainty
The talk is devoted to the data fitting problem which requires constructing a functional dependence, of a given type, that best fits a set of inaccurate data. Often, this problem is also called regression analysis problem.
The specificity of the formulation we are considering is that the inaccurate data have interval uncertainty, i.e. we know only the lower and upper bounds of their possible values, while the probabilistic characteristics of the data errors are absent. It is shown that the problem under consideration reduces to solving interval systems of equations constructed from the processed data. The talk surveys various approaches to the solutions of the interval data fitting problem, the paradoxical nature inherent in its very formulation, as well as possible ways to overcome it. We propose a promising technique, called maximum compatibility method, for the solution of the data fitting problem under interval uncertainty, and then consider its practical implementation for the case of the linear function. Finally, properties of function parameter estimates obtained by the maximum compatibility method are discussed.
Sergey P. Shary (in Russian — Сергей Петрович Шарый) is a professor and head of Mathematical Modeling Department at Novosibirsk State University, Russia.
Born in 1962, Semipalatinsk, USSR, he graduated from Novosibirsk State University, Novosibirsk, Russia, with a M.S. degree in pure and applied mathematics in 1985. PhD thesis in computational mathematics was defended in 1992, and in 2000 he received a doctorate in the same specialty.
In 1985-1995, he worked at the Computing Center of the Russian Academy of Sciences in Krasnoyarsk, and from 1996 to the present, he has been working at the Institute of Computational Technologies in Novosibirsk and in Novosibirsk State University. In 1997-2007, he worked concurrently in the IT industry, first as a consultant to the Interval project at Sun Microsystems, and then as a senior software engineer at Intel Corporation (Novosibirsk site).
In the fall semester of 2000, he was a visiting professor at the Faculty of Informatics and Applied Mathematics in University of Girona, Spain.
The research interests include interval analysis, computational mathematics, mathematical programming and data analysis. Over 150 research articles published.
Homepage — http://www.nsc.ru/interval/shary
The MICELab group appreciates his visit and the interesting talk.