search

ibex-team / ibex-ampl

Plugin for AMPL modeling language
GNU General Public License v3.0
1 stars 0 forks source link

AMPL bug when reading file s365mod.nl #2

Open bneveu opened 5 years ago

bneveu commented 5 years ago

The variables x(2]nd x(3) are exchanged when reading the file s365mod.nl (in the coconut series2 benchmark) The created system in ibexopt does not correspond to the input file in particular the objective becomes goal: (x(1)x(2)) instead of (x(1)x(3)) and the constraints constraints: (((x(5)-x(7))^2+(x(4)-x(6))^2)-4)>=0 ((((x(2)x(4))-(x(3)x(5)))/(x(2)^2+sqrt(x(3)^2)))-1)>=0 ((((x(2)x(6))-(x(3)x(7)))/(x(2)^2+sqrt(x(3)^2)))-1)>=0 ((((((x(3)-x(1))x(5))+(x(1)x(2)))-(x(2)x(4)))/((x(3)-x(1))^2+sqrt(x(2)^2)))-1)>=0 ((((((x(3)-x(1))x(7))+(x(1)x(2)))-(x(2)x(6)))/((x(3)-x(1))^2+sqrt(x(2)^2)))-1)>=0

instead of constraints: (((x(4)-x(6))^2+(x(5)-x(7)^2)-4)>=0 ((((x(3)x(4))-(x(2)x(5)))/(sqrt(x(2)^2)+x(3)^2))-1)>=0 ((((x(3)x(6))-(x(2)x(7)))/(sqrt(x(2)^2)+x(3)^2))-1)>=0 (((((x(1)x(3))+((x(2)-x(1))x(5)))-(x(3)x(4)))/(sqrt(x(3)^2)+(x(2)-x(1))^2))-1)>=0 (((((x(1)x(3))+((x(2)-x(1))x(7)))-(x(3)x(6)))/(sqrt(x(3)^2)+(x(2)-x(1))^2))-1)>=0

bneveu commented 5 years ago

The ampl generator of .nl files may change the variable order : it is then a little complicated to bound them manually and to check the solution. I think it is better to let the unbounded variables in the problem definition unbounded in the .nl file and to bound them if necessary at the beginning of the optimization process.

gchabert commented 5 years ago

Good to know! Thanks Bertrand

Jordan08 commented 5 years ago

Hello, I just saw the bug. Indeed, AMPL chooses the order of the variables. The problem does not come from this reordering. When the bounds are in the model, they are well affected. The problem is that variable names are not keeping. It is necessary to modify the interfaceAMPL to recover the names of the variables from the model AMPL. (I did not do it because I do not know how to do it yet)

If we put the bounds outside the model, it will be problematic. Because some bounds are given in the modeling like constraints, and they are interpreted by AMPL as bounds. In fact, the bounds are really part of the model. If we evacuate them from the model, the risk is to lose information (it has already happened to me ....)

To investigate ... from where the reopening of the bug ...