%A Edward J. Anderson
%A Michael C. Ferris
%T A Direct Search Algorithm for Optimization with Noisy Function Evaluations
%D November 1996
%R 96-11
%I COMPUTER SCIENCES DEPARTMENT, UNIVERSITY OF WISCONSIN
%C MADISON, WI
%X
We consider the unconstrained optimization of a function when each
function evaluation is subject to some random noise. Our model of
computation assumes that averaging repeated observations at the same
point can lead to a better estimate of the underlying function value.
In practice problems of this form may occur when choosing the best
settings for the controls in a processing plant, or in choosing the
parameters in an experiment of some kind. We consider direct search
methods with the possibility of repeated function evaluations at the
same point. We describe an algorithm of this type which has reasonable
computational performance and for which convergence can be
established.