Guillaume Marques
Guillaume Marques
Hi, Thank you very much for this detailed description of the bug. > Of course in the latter case one could simply use the Simplex method, but the model of...
Hi, > **Integrality on columns** > > If I understand your suggestion and the consequences in Coluna correctly, it would mean that as soon as there is an integer variable...
I reopen beause #551 does not address the whole issue.
Broken tests because of JuMP 1.2.0 release. Update: fixed in https://github.com/atoptima/BlockDecomposition.jl/pull/89
Hi, You can indeed define a time limit for your subsolver using an `optimizer_with_attributes` from JuMP : https://github.com/atoptima/Coluna.jl/blob/7a7d72688e3b326d0ad7273b38cd5adb828a9785/test/optimizer_with_attributes_test.jl#L13 These parameters will be applied to all subsolvers.
Yes, it restricts the execution time for each subsolver call. Unfortunately, there is no way to define a time limit on the whole execution of Coluna at the moment. It...
Hey Ruslan, Thanks for the report, I'll take a look tomorrow.
`x` variables are unbounded. So when x variables have negative costs in the subproblem, the subproblem objective is unbounded and no column is generated. Col gen ends up by saying...
If I add a lower bound on x variables (x[] >= 0), I find the same solution with GLPK and Coluna. direct approach with GLPK ``` JuMP.objective_value(model) = 64.14285714285715 x[1,1]...
> Ok, thank you, Guillaume. I think however that if the subproblem is unbounded, the master solution should also be declared as unbounded, no? It is not correct to declare...