We report the development of second-derivative FORTRAN routines to sup
plement Algorithm 566 developed by J. More et al. (ACM Trans. Math. So
ftw. 7, 14-41, 136-140, 1981). Algorithm 566 provides function and gra
dient subroutines of 18 test functions for multivariate minimization.
Our supplementary Hessian segments enable users to test optimization s
oftware that requires second derivative information. Eigenvalue analys
is throughout the minimization is now possible, with the goal of bette
r understanding progress by different minimization algorithms and the
relation of progress to eigenvalue distribution and condition number.