Re: 'sqp' error state equal to 0 - what does it mean?

[Guido Walter Pettinari]

Hi John,

thank you for looking at the function.

I am confident that my starting point does not satisfy the optimality conditions
since (i) the minimum I get from 'sqp'  is different from the starting point and
(ii) the computation takes several seconds.

Cheers,

Guido


On Apr 14, 2010, at 14:37 , John W. Eaton wrote:

> On 14-Apr-2010, Guido Walter Pettinari wrote:
> 
> | Thank you for your answer, Francesco.
> | 
> | I am sorry if I was not clear enough. I am aware that info=101 is good, my 
> concern
> | is that I am getting a result (info=0) which is not contemplated by the 
> function
> | documentation.
> 
> Taking a quick look at sqp.m, I think this can happen if the following
> conditions are all satisfied on the initial iteration.
> 
>      t0 = norm (c - A' * lambda);
>      t1 = norm (ce);
>      t2 = all (ci >= 0);
>      t3 = all (lambda_i >= 0);
>      t4 = norm (lambda .* con);
> 
>      if (t2 && t3 && max ([t0; t1; t4]) < tol)
>       break;
>      endif
> 
> So, if this is correct, then I think your starting point satisfies the
> optimality conditions.
> 
> jwe