Hi team, I'm using GUROBI to solve a large-size MPS file which is roughly about 60G. This problem contains
26423572 rows, 265646022 columns, 962620739 nonzeros
I tried with three methods, which are respectively: barrier, dual simplex, and primal simplex.
Both barrier and primal simplex gives us the similar optimal value:
Optimal objective -1.550697119e+02
However, dual simplex gives the objective value of
Optimal objective -1.467550711e+02
This is too off from the true opt value, meaning that the dual simplex method has something wrong. In fact, in the last few iterations, the output log file of dual simplex reports significantly positive Dual Inf. value, which is contradictory to the fact.
May I ask why the dual simplex method gives the wrong solution?
Please sign in to leave a comment.