ILP Mappings and Schedule

Given a sufficiently simple set of ML models, an ILP could solve for an optimal mapping. ILPs are known for increasing their computational time exponentially with the number of required variable for a given problem. As such, an ILP solver might only be applicable for smaller problems and have to give way to a heuristic search when a problem is scaled too large.

The ILP implementation found in this work uses the Gurobi optimization framework for solving the formulated ILP problem. The ILP formulation itself can be found in this PDF.

Given that the ILP solved the entire problem, the OpenDSE heuristic framework used for a large part of this work can be completely skipped as there is no need for any heuristic searching.