prerequisites

• gurobi (and gurobipy)
• python 3
• numpy
• pandas
• matplotlib
• sbt
• scala
• java

Data

The Canadian dataset is contained in the folder CanadianInstances The US dataset is contained in the folder PortoInstances The type information files containt information relative to the PRA

The cleaned up graphs are contained in the folder src/data These are the files used to run the experiments

CP implementation

The code for the CP implementation is contained in the folder all_solutions

MIP implementation

The code for the MIP implementation is contained in the folder src

List of experiments and files

1. Enumerate all solutions

   ./all_solutions/propagator/oscar-propagator/target/pack/bin/enum-kep --kep-file input-file [--cycle-limit 3] [--time-limit time] [--LP-prop bool] [--edge-prop bool] [--output-file filename]

2. Enumerate all solutions hierarchical

   ./all_solutions_hierarchical/propagator/oscar-propagator/target/pack/bin/enum-kep --kep-file input-file [--cycle-limit 3] [--time-limit time] [--LP-prop bool] [--edge-prop bool] [--output-file filename]

3. Enumerate all solutions relaxed

   python src/kep_mip_relaxed.py input-file output-file

4. Find projected solutions

   python src/experiments/all_solution_cp/preprocess.py [input-file] [output-file]

5. Find the alpha value for instances

   python src/alpha.py --loss=l2

6. Runtime plot (Figure 3)

   ipython src/analysis/plot_performance_profiles.ipynb

7. Number of solutions (table 2)

   python src/experiments/all_solution_cp/num_solutions.py

8. Loss plots (figure 4)

   python src/experiments/all_solution_cp/plot.py

9. Alpha and OPT values for instances (table 3)

   python src/alpha.py

10. New patients after relaxation (Figure 5)

    python src/new_patients.py