[Ipopt-tickets] [Ipopt] #85: C Interface enhancements for global optimization

[Ipopt]

#85: C Interface enhancements for global optimization
------------------------------+---------------------------------------------
Reporter:  sergm              |       Owner:  ipopt-team
    Type:  enhancement        |      Status:  new       
Priority:  normal             |   Component:  Ipopt     
 Version:  3.5 (C++ Version)  |    Severity:  normal    
Keywords:                     |  
------------------------------+---------------------------------------------
 Dear colleagues,

 We try to use “Ipopt” in global optimization. Our algorithm based on
 branch and cut technique and use “Ipopt” for solving relaxation problems
 in the nodes of the search tree. We use “C Interface” because our program
 written on Visual Basic. This interface is not yet suitable for our
 purpose. The minimal requirements are follows:

 1. “IpoptSolve” returns a lot of codes. Practically in branch and cut
 algorithm those codes are useless. In really only 4 codes are required for
 those purposes: “optimal solution found”, “feasible solution found”,
 “problem is infeasible”, “unknown”. For example, if “Ipopt” found the
 feasible solution and after that the “out of memory problem” happened then
 result code should be “feasible solution found”.

 2. “C Interface” has no the branching supporting. Today we force
 recreation the “Ipopt” object for each relaxation problem. The convenient
 interface should have the special functions for setting variables bounds
 and activation/deactivation of constraints. With this functions variables
 and constraints logic (OR, AND, IMP and so on) can be branched.

 3. “C Interface” would have a callback functions which are raise on
 feasible (or optimal) solution found, on algorithm user break request and
 on log write.

 Very interesting issue is the conditions on which we can believe the
 “optimal” and “infeasible” results of “Ipopt”. If the Hessian is non-const
 then can we believe low (high) bounds of interval computations in
 convexity (concavity) test?

-- 
Ticket URL: <https://projects.coin-or.org/Ipopt/ticket/85>
Ipopt <http://projects.coin-or.org/Ipopt>
Interior-point optimizer for nonlinear programs.