spopt
spopt copied to clipboard
potential for scipy.optimize.mip
scipy.optimize
now offers mixed integer programming solution methods through a HiGHS interface. This is currently in the main
branch and will be part of the v1.9.0
release, which should be cut soon. We may want to see about incorporating this new functionality due to the COIN-OR funding situation.
It seems a good tool. Maybe we should do an example for a simple locate model like LSCP? Then we can discuss about API changes and other enhancements to adopt that API. What do you think?
Seems like a very good open-source project from Edinburgh! I think we can test the speed, compare these 2 methods, and see what we want to do.
pulp
now supports HiGHS. So now we can easily implement that solver in a demo/example. And if we decide to support our MIPs through scipy.optimize
with HiGHS later, it will provide an easier transition for testing and ensuring equivalent solutions are obtained.
I have have successfully installed HiGHS and used from the pulp
interface. The demo runs will be in the facloc_real-world_*
notebooks and added in #300.
