Error message, couenne solver
Mario
I'm working with a mixed integer nonlinear model, I use the solver
couenne but the solution appears in the following error message:
Error in an AMPL evaluation. Run with "halt_on_ampl_error yes" to see
details.
Error in an AMPL evaluation. Run with "halt_on_ampl_error yes" to see
details.
Cbc0012I Integer solution of 54.8112 found by Couenne Rounding NLP
after 0 iterations and 0 nodes (0.00 seconds)
Cbc0006I The LP relaxation is infeasible or too expensive
Cbc0013I At root node, 0 cuts changed from Objective-9.99e-9.99e +12
to +12 in 1 passes
Cbc0014I Cut generator 0 (Couenne convexifier cuts) - 0 row cuts
average 0.0 elements, 2 column cuts (2 active)
Cbc0001I Search completed - best 54.81118681014511 Objective, Took 0
iterations and 0 nodes (0.03 seconds)
Cbc0035I Maximum depth 0, 0 variables fixed on reduces cost
however, the solver generates a solution
I do not know why this error occur?, This solution is reliable?
Thank you for the information you could provide me.
Robert Fourer
constraint function being uncomputable at some iterate that has been reached
by the solver algorithm. For example the solver may have reached (or
started at) a point where some function has a division by zero, or takes the
derivative of a square root at zero, or the square root of a negative
number. AMPL tells the solver that a function cannot be evaluated at such a
point, but it is up to the solver to decide how to handle or report such a
situation.
Since this is specifically a problem with the Couenne solver, you may need
to report it directly to the developers of Couenne, as explained at
https://projects.coin-or.org/Couenne.
Bob Fourer
4...@ampl.com
> --
> You received this message because you are subscribed to the Google
> Groups "AMPL Modeling Language" group.
> To post to this group, send email to am...@googlegroups.com.
> To unsubscribe from this group, send email to
> ampl+uns...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/ampl?hl=en.
Mario Basallo
Pietro Belotti
messages are from a component used by Couenne, Ipopt. Ipopt is used to
compute an initial feasible MINLP solution, if any exists. Couenne is
able to start the search even if no solution is found or if such
errors arise. The solution provided by couenne, therefore, is optimal
regardless of those errors.
Pietro
On Apr 13, 9:48 pm, "Robert Fourer" <4...@ampl.com> wrote:
> Typically an "Error in an AMPL evaluation" is due to an objective or
> constraint function being uncomputable at some iterate that has been reached
> by the solver algorithm. For example the solver may have reached (or
> started at) a point where some function has a division by zero, or takes the
> derivative of a square root at zero, or the square root of a negative
> number. AMPL tells the solver that a function cannot be evaluated at such a
> point, but it is up to the solver to decide how to handle or report such a
> situation.
>
> Since this is specifically a problem with the Couenne solver, you may need
Robert Fourer
As far as I know, there is no solver that will try to generate a list of all feasible solutions for a nonlinear problem (assuming all variables are integer, so that the list is finite). If the constraints are all linear then you could use the CPLEX facility for generating a given number of optimal solutions, as described recently in this group (search "multiple optimal solutions").
In special cases for very small problems, you can write an AMPL script using some loops to generate all possible combinations of variables and to check them for feasibility -- I have done such things to create examples used in classes -- but I can't give a general explanation as the details are very specific to each problem.
Bob Fourer
From: am...@googlegroups.com [mailto:am...@googlegroups.com]
On Behalf Of Mario Basallo
Sent: Wednesday, April 13, 2011 4:58 PM
To: am...@googlegroups.com
Subject: Re: [AMPL 4556] Error message, couenne solver