Tobias Achterberg

Gurobi accepts all SOC in standard form. But when it comes to rotated or affine SOCs, it may not be able to detect the structure. Then, it falls back to the regular Q constraint mechanism, which re...

A constraint x1 + ... + xn <= 1 with xj being binary variables says that at most one them can be non-zero. This is equivalent to forming the SOS1 constraint SOS1(x1,...,xn). But the SOS1 constraint...

Sorry, this I don't know. You would need to ask in the PULP forums. I don't know the syntax of PULP or PyCharm.  

As far as I can see, you are minimizing your objective function. Hence, the primal (incumbent) bound is an upper bound, while the dual bound is a lower bound. You can provide a "MIP start" solution...

So, what happens with different seeds? Do you see that independent on the seed, Windows is slower than Mac? When you look at the Windows Task Manager and observe that only 20% of the CPU is used, i...

Hi Henrique, you could install a callback function that queries the MIP_NODLFT data. The first time this becomes non-zero, you are at the end of the root node. Alternatively, you could launch your ...

As I said, this could just be due to performance variability (i.e., good or bad luck with the heuristic decisions that the solver takes). For example, it might be that on Mac Gurobi decides more qu...

This is really strange. Your models are not that large that they would lead to 300 seconds just for solving an equation system. Yes, could you please send your data to achterberg (at) gurobi (dot) ...

I can see that between 1200 and 2500 seconds, the node throughput is pretty low. This can happen, because at some points during the search we occasionally apply very heavy algorithms with cutting p...

No, there are no parameter settings that will *certainly* make the process faster. It is always a matter of trying. And if you want to systematically try parameter combinations you should take a lo...