Abstract

DOGLEG METHOD WITH BROYDEN CLASS UPDATING TECHNIQUE FOR SOLVING TRUST-REGION SUB-PROBLEMS OF SOME UNCONSTRAINED MULTIVARIATE NONLINEAR OPTIMIZATION PROBLEMS

E. C. Duru1 and B. I. Oruh2

---

In this paper, the Dogleg-type trust-region method that employed the Broyden Class updating techniques in generating the approximation matrices to the hessian of the objective function is presented where convergence is based on constructing two paths. The conditions on the paths were incorporated into the algorithm used in determining the optimum points of the smooth functions considered. Numerical computations on some test functions showed that this procedure is efficient and globally convergent. The result equally highlighted the effect of the Broyden class parameter on the convergence of the solution. Keywords: Trust-region methods, Dogleg method, Broyden class updating technique.

PDF