Article

Article title CONDITIONALITY OF PARAMETRIC OPTIMIZATION
Authors V.N. Biryukov
Section SECTION IV. MATHEMATICAL METHODS OF AN ARTIFICIAL INTELLECT
Month, Year 06, 2012 @en
Index UDC 621.396:517.9:518.6
DOI
Abstract If the accuracy of a source data is limited, the error of parametric optimization is random. The error depends on the stiffness of a problem, and on the conditioning of the argument. This paper provided opportunity to experimental evaluation of the error. Component of the error associated with the high stiffness of the problem; in some cases it may be negligible. In these cases, the error component associated with poor conditioning becomes dominant. The article shows that for low accuracy of initial data ill-conditioning becomes a major factor of optimization error. The most important conclusion relates for a multi-dimensional problems, since the probability of ill-conditioning increases with increasing dimension.

Download PDF

Keywords Index Terms-optimization; error; conditionality.
References 1. Ракитский Ю.В., Устинов С.М., Черноруцкий И Г. Численные методы решения жестких систем. – М.: Наука, 1979. – 211 с.
2. Press W.L. et al. Numerical recipes: the art of scientific computing. Cambridge university press, 1988. - 620 с.
3. Бирюков В.Н. Жесткие задачи радиоэлектроники // Деп. ВИНИТИ, № 196 – В 2007 от 05.03.07 - 120 с.
4. Бирюков В.Н. Оценка точности определения параметров моделей полевого транзистора // Известия вузов – Электроника. – 2010. – Т. 15. – № 4. – С. 22-27.
5. Федоренко Р. П. Введение в вычислительную физику. – М.: Изд-во Моск. физ.-техн. ин-та, 1994. – 528 с.

Comments are closed.