? non oredr-2 obj function

Hi:

2nd order objective functions were used quite often in optimization

due to the fact that one can then solve linear equations to determine

critical point(s) satisfying the necessary condition for optimization.

However, in terms of sensitive of objective function, isn't it true

that higher order objective functions are more sensitive? Besides,

when one encounters nonlinear constraints, it becomes meaningless

to keep that 2nd order objective function, since even after
differentitation,

one still has those nonlinear equaions from constraints.

Thus, my question is: are there already some evidences in theory or

from experience that non 2nd order objective functions serve better

for some kinds of optim problems?

BTW, if one is sure that the objective function value is never zero, will

square-root serves as good as a quadratic func since they are symmetric

to y = x, so convergence rates (or gradient) are the same?

Thanks,
by Cheng Cosine
Mar/30/2k8 NC


.