[SciPy-User] Sometimes fmin_l_bfgs_b tests NaN parameters andthen fails to converge

[Steven G. Johnson]

On Jan 4, 3:23 pm, josef.p... at gmail.com wrote:
> If you have local maxima, and your optimization problem is not well
> behaved, then you might need a global optimizer.
> I think also the algorithms innloptwill get stuck at local optima,
> since all these optimizers only use local information.

NLopt also includes global optimization algorithms.  (Of course,
global optimization problems can be NP hard, so there is no magic
bullet in general.)