A pattern search method guided by simplex derivatives for use in derivative-free optimization
SID-PSM is a solver for constrained or unconstrained nonlinear optimization problems, using derivative-free
methods. In the general constrained case and for the current version,
the derivatives of the functions defining the constraints must be provided. The optimizer uses an implementation
of a generalized pattern search method, combining its global convergence properties with the efficiency
of the use of quadratic polynomials to enhance the search step
and of the use of simplex gradients for guiding the function evaluations of the poll step.
SID-PSM is freely available for research,
educational or commercial use, under a GNU lesser general public license.
- A. L. Custódio and L. N. Vicente, Using sampling and simplex derivatives in pattern search methods,
SIAM Journal on Optimization, 18 (2007) 537-555 PDF (complete numerical results)
- A. L. Custódio, H. Rocha, and L. N. Vicente, Incorporating minimum Frobenius norm models in direct search,
Computational Optimization and Applications, 46 (2010) 265-278 PDF
The data profiles for the solvers NMSMAX and SID-PSM
on the stochastic noisy test problems are slightly different
from the true ones. Those corrected profiles can be found
and do not affect any of the conclusions of the paper.
The SID-PSM team:
Ana Luísa Custódio (New University of Lisbon)
Luís Nunes Vicente (University of Coimbra)