Prev: Problem with Data extraction from multidimensional arrays
Next: Finding Gaussian Basis
From: Plucesiar on 14 Jun 2010 22:43
Hello, I have a question about fmincon. If the user-supplied starting point is infeasible, but assuming the problem is feasible, what does fmincon do? Does it try to find the nearest point that's feasible without consideration of the objective function?

Thanks
wly
From: Alan Weiss on 15 Jun 2010 08:42
On 6/14/2010 10:43 PM, Plucesiar wrote:
> Hello, I have a question about fmincon. If the user-supplied starting
> point is infeasible, but assuming the problem is feasible, what does
> fmincon do? Does it try to find the nearest point that's feasible
> without consideration of the objective function?
>
> Thanks
> wly

Essentially, for bounds and linear constraints, indeed fmincon
active-set and SQP algorithms ignore the objective function and solve a
linear programming problem to obtain a feasible point. They use a merit
function for nonlinear constraints.

For a complete description, see
http://www.mathworks.com/access/helpdesk/help/toolbox/optim/ug/brnoxzl.html#brnox01

Look at the section heading "Initialization"

Alan Weiss
MATLAB mathematical toolbox documentation
 | 
Pages: 1
Prev: Problem with Data extraction from multidimensional arrays
Next: Finding Gaussian Basis