Abstract

The determination of the global extremum of a function is a notorious numerical problem when there are local extrema present. The numerical algorithm which has to determine the global extremum iteratively is in this case very likely to produce a local extremum in the neighborhood of the initial guess of the solution. Moreover, usually one cannot be sure not to have missed an extremum, i.e. various procedures together with various initial values might still overlook the global extremum.

© 1983 Optical Society of America