Prev: What algorithm behind ppual() ? : Evaluate ppform function.
Next: merging ODE45 results
From: MRINMAY DEB on 19 Apr 2010 12:56
Hi All,
I am using 'active-set' algorithm in fmincon to optimize a multidimensional nonlinear continuous function with only bound constraints. In the output, I get the message "Local minimum possible, constraints satisfied" and exitflag is 5. This means that
"the predicted change in the objective function,
is less than options.TolFun, and the maximum constraint
violation is less than options.TolCon". Fine!

I observe that the firstorderopt is of the order e-008 and directional derivative is of the order e-007, however, the 2-norm of the gradient at the stopping point is pretty high (~5). Can anyone please explain why does it happen so that although first order optimality value and directional derivative is zero, the gradient is not. Does this mean fmincon didnt find global min? How to overcome this problem?
Any help will be appreciated.

Thank you in advance
Mrinmay
 | 
Pages: 1
Prev: What algorithm behind ppual() ? : Evaluate ppform function.
Next: merging ODE45 results