[Ipopt] The difference between the current implementation of Ipopt and the 2005 Math Prog A paper

[Stefan Vigerske]

Hi,

Ipopt internally still solves the problem given in the paper.
The TNLPAdapter class takes care of reformulating the commonly used 
general form with inequalities (TNLP) into the formulation with 
equalities only by introducing slack variables.

Stefan

On 07/14/2013 07:58 PM, Peng Yu wrote:
> Hi,
>
> The current ipopt document says that ipopt solves the following problem.
>
> \min_{x∈R^n} f(x)
> s.t. g_L ≤ g(x) ≤ g_U
> x_L ≤ x ≤ x_U
>
> But the 2005 Math Prog A paper solves the following problem.
>
> \min_{x∈R^n} f(x)
> s.t. c(x) = 0
> x_L ≤ x ≤ x_U
>
> So there is a difference between the two. The paper only described the
> equality constraints but the current implement can solve the
> inequality constraints. Does anybody know whether there is a paper
> describes all the current implementation details? (If not, what part
> in the paper is still relevant to the current implementation and what
> part is not relevant any more?) Thanks.
>