Distributed algorithms and hierarchic optimization for solving a parameter inverse problem

• Autorzy: Telega, H.
• Języki publikacji: EN

Abstrakty

EN

In this paper we present some numerical algorithms for the identification of parameters in inverse problems. We are motivated by the problem of identification of physical parameters in the Swartzendruber formula which appears in the mathematical model of the prelinear filtration of ground water. The parameter inverse problem is formulated as an appropriate control one. An application of standard nonlinear programming methods for solving the problem is presented and obstacles of this approach are pointed out. We propose three strategies in order to deal with many local minima and sensitivity of the algorithms on starting points. The first one is to improve local minimum search in large parts of the domain and the third one is to fasten computations using parallel and distributed techniques. Two algorithms that correspond to these strategies are presented: the hierarchic optimization algorithm and the algorithm of simple parallel search with distribution of computations (multiple starting point parallel optimization). Some modifications to these algorithms are also proposed. The results of the application of the described algorithms to a tested problem are presented. The The computational tests are based on real data obtained from an experimental range.

Słowa kluczowe

EN

inverse problems; hierarchic optimization; parallel distributed algorithms

• Czasopismo: Zeszyty Naukowe Uniwersytetu Jagiellońskiego. Prace Informatyczne
• Rocznik: 1998
• Tom: z. 8
• Strony: 7-28

Twórcy

autor

Telega, H.

• Instytut Informatyki Uniwersytet Jagielloński ul. Nawojki 11 30-072 Kraków,