Пожалуйста, авторизуйтесь:
ФГОУ ВПО Сибирский федеральный университет
Российская Федерация, 660041, г. Красноярск, пр. Свободный, 79
|
|
Michael
professor, doctor of physics and mathematics (University of North Carolina at Charlotte, Charlotte, North Carolina, USA, professor)
|
Секция: Inverse problems and ill-posed problems of mathematical physics
Тема доклада:
|
|
"A globally convergent numerical method for some
coefficient inverse problems
" |
Тезисы доклада:
|
|
The problem of multiple local minima of residual least squares functionals plagues the development of reliable numerical methods for coefficient inverse problems. As a result, all current algorithms are locally convergent ones. That is they require a good first guess for their solution. This is unsatisfactory for applications. In this talk we will present a new development: a globally convergent numerical method for a wide class of coefficient inverse problems. This algorithm does not require a good first guess. It avoids the construction of least squares objective functionals.
|
|
|