1

I am trying to implement an incremental optimization where I can feed my MILP model with new constraint/variables and remove some other constraints/variables as time passes. Not to mention solved variables should be treated as fixed values as long as they are not removed.

My question is if there is a standard way/tools to do it? What I thought about so far is to manipulate GAMS sqlite database using python, any better idea? It is fine to use other environments.

I appreciate any clue to implement such thing. Thanks

behzadb
  • 25
  • 1
  • 5
  • If you drop GAMS and just use any (more low-level?) modelling-tool, like gurobipy (which is Gurobi's python-interface) for example, i don't see a problem. There might be some more theoretical and practical issues though (full restart vs. mip-start vs. warm-start vs. hot-start and so on) which is probably somewhat solver-specific (and sometimes it might help to know more of ones own solver internals). Read [this recent answer too](https://stackoverflow.com/a/50831513/2320035). (i never used GAMS and can't evaluate the possibilities there) – sascha Jun 24 '18 at 10:52
  • There are different ways to interpret this question. I think you are asking whether model equations in GAMS can be made "data driven". The answer is (obviously) yes. This means the number of equations is not "static" but depends on the data. When new data arrives, different equations are generated. – Erwin Kalvelagen Jun 26 '18 at 02:24

0 Answers0