infeasible with secondary target #5
Description
Hi,
I really appreciate this work and GAPSPLIT helped us to achieve sampling of a heavily constrained MILP model that cannot be done before. However, I can only use gapsplit without secondary targets. allowing secondary targets will make the model infeasible.
after a quick check of the code, it looks line 231 in gapsplit.m appears weird:
model.A(kridx,:) = 0;
this line will erase the whole row so that the additional variables for secondary targets are essentially masked. It looks this will further impose hard constraints on the flux of the secondary targets, such that the model is infeasible. I guess the bug-free code should be
model.A(kridx,~kcidx) = 0;
I am wondering if this is an accidental bug when you update the github. maybe I am being naive, but I am happy to discuss more.
Thanks!
Hang