README.ibexopt

IbexOpt is a solver based on IBEX versions 2.8.9 that can be used
either "stand-alone" or with AMPL's options to solve nonlinear 
problems expressed in AMPL.

To use IbexOpt with AMPL, you have several options.  You can invoke
it within an AMPL session by saying

    solve;

or, if $solver is not already ibexopt,

    option solver ibexopt;
    solve;

When $solver has its default value (ibexopt), AMPL's solve command invokes

    ibexopt stub -AMPL

If a stub is present, IbexOpt tries to write the computed solution to
stub.sol

For details not given here, see the IBEX documentation:
http://www.ibex-lib.org

Source for IBEX is available from
https://github.com/ibex-team/


-----------------------
solve_result_num values
=======================

Here is a table of solve_result_num values that "ibexopt" can return
to an AMPL session, along with the text that appears in the associated
solve_message.

 Value  Message

    0   OPTIMIZATION SUCCESS! 
        The global minimum (with respect to the precision 
        required) has been found. In particular, at least 
        one feasible point has been found, less 
        than obj_init_bound, and in the time limit.
    200 INFEASIBLE PROBLEM.
        No feasible point exist less than initial_loup.
        In particular, the function returns INFEASIBLE 
        if the initial bound \"obj_init_bound\" is LESS 
        than the true minimum (this case is only possible
        if abs_eps_f and rel_eps_f are 0). In the
        latter case, there may exist feasible points.
    201 NO FEASIBLE POINT FOUND. 
        No feasible point could be found less than 
        obj_init_bound. Contrary to INFEASIBLE, 
        infeasibility is not proven here. Warning: this 
        return value is sensitive to the abs_eps_f and 
        rel_eps_f parameters. The upperbounding makes the 
        optimizer only looking for points less than 
        min{(1-rel_eps_f)*obj_init_bound,obj_init_bound-abs_eps_f}.
    300 UNBOUNDED OBJECTIVE FONCTION. 
        The objective function seems unbounded (tends to -oo).
    400 TIMEOUT.
        time limit is reached.
    402 UNREACHED PRECISION. 
        The search is over but the resulting interval [uplo,loup]
        does not satisfy the precision requirements. There are
        several possible reasons: the goal function may be too 
        pessimistic or the constraints function may be too 
        pessimistic with respect to the precision requirement 
        (which can be too stringent). This results in tiny boxes 
        that can neither be contracted nor used as new loup candidates. 
        Finally, the eps_x parameter may be too large.
        

-----------------------
ibexopt option
=======================

To set these directives, assign a string specifying their values to the AMPL option ibexopt_options. For example:

    option ibexopt_options 'rel_eps_f=1e-5 timeout=1000 rigor=1 simpl_level=2';


Keywords are followed by a value except those marked No value is expected in the listing.

 rel_eps_f      Relative precision on the objective. Default: 1e-3. 
 abs_eps_f      Absolute precision on the objective function. Default: 1.e-7. 
 eps_h          Relaxation value of the equality constraints. Default: 1.e-8. 
 random_seed    Random seed (useful for reproducibility). Default: 1.  
 timeout        Timeout (time in seconds). Default: -1 (none). 
 obj_numb       Choose which objective function of the AMPL model: 0 = none, 1 = first. Default: 1.
 simpl_level    Expression simplification level. Possible values are:
                0 : no simplification at all (fast).
                1 : basic simplifications (fairly fast). E.g. x+1+1 --> x+2
                2 : more advanced simplifications without developing (can be slow). 
                    E.g. x*x + x^2 --> 2x^2 
                3 : simplifications with full polynomial developing (can blow up!).  
                E.g. x*(x-1) + x --> x^2 
                Default value is : 1.
 init_obj_value Initialization of the upper bound of the objectif function
                with a known value. Default: +infinity. 
 rigor          Activate rigor mode (certify feasibility of equalities). Default: -1.
                 -1 : By default: #default_rigor.
                  0 : Deactivate rigor mode
                  1 : Activate rigor mode, feasibility of equalities is certified.
 kkt            Activate contractor based on Kuhn-Tucker conditions. Default: -1.
                 -1 : By default, we apply KKT only for unconstrained problems.
                  0 : deactivate KKT contractor
                  1 : activate KKT contractor
 inHC4          Activate feasibility search with LoupFinderInHC4. Default: -1.
                 -1 : By default: #default_inHC4.
                  0 : Deactivate inHC4
                  1 : Activate inHC4, feasibility is also tried with LoupFinderInHC4.
 trace          Activate trace. Updates of lower and upper bound are printed while minimizing. Default: 1. 
                  0 : nothing is printed
                  1 : prints every loup/uplo update.
                  2 : prints also each handled node (warning: can generate very long trace).

-------------------------------