A note on two fixed point problems (Q2465828)

The authors discuss and extend the applicability of the exterior ellipsoid algorithm for approximating the fixed points of a system of nonlinear equations. It is shown that the ellipsoid algorithms could be extended and applied to larger classes of functions that are only nonexpanding in the direction of fixed points. These larger classes contain functions that may be globally expanding, may be noncontinuous or may have unbounded derivatives. Also, the authors are focused on the design of a nearly optimal algorithm that is applied to univariate fixed point problems originating in modeling combustion of energetic materials. The fixed points calculations are utilized in large scale transient combustion simulation. They are repeated at every grid cell of a very large model and at every time step. This is why they have to be extremely fast and sufficiently accurate. An almost optimal (on the average) hyper-bisection/secant modification of a hybrid bisection-regula falsi-secant method, to solve a nonlinear fixed point problem is derived. A significant improvement in the cost of calculation, by the reduction of the average number of iterations, for the specific univariate zero finding problem, are obtained.