Applying the pattern search implicit filtering algorithm for solving a noisy problem of parameter identification
| Autor: | M. A. Diniz-Ehrhardt, D. G. Ferreira, Sandra A. Santos |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
021103 operations research
Control and Optimization Applied Mathematics 0211 other engineering and technologies Ode 010103 numerical & computational mathematics 02 engineering and technology Lipschitz continuity 01 natural sciences Pattern search Parameter identification problem Computational Mathematics Noise Ordinary differential equation Convergence (routing) Derivative-free optimization 0101 mathematics Algorithm Mathematics |
| Zdroj: | Computational Optimization and Applications. 76:835-866 |
| ISSN: | 1573-2894 0926-6003 |
| DOI: | 10.1007/s10589-020-00182-2 |
| Popis: | Our contribution in this paper is twofold. First, the global convergence analysis of the recently proposed pattern search implicit filtering algorithm (PSIFA), aimed at linearly constrained noisy minimization problems, is revisited to address more general locally Lipschitz objective functions corrupted by noise. Second, PSIFA is applied for solving the damped harmonic oscillator parameter identification problem. This problem can be formulated as a linearly constrained optimization problem, for which the constraints are related to the features of the damping. Such a formulation rests upon a very expensive objective function whose evaluation comprises the numerical solution of an ordinary differential equation (ODE), with intrinsic numerical noise. Computational experimentation encompasses distinct choices for the ODE solvers, and a comparative analysis of the most effective options against the pattern search and the implicit filtering algorithms. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
| Nepřihlášeným uživatelům se plný text nezobrazuje | K zobrazení výsledku je třeba se přihlásit. |