How to reproduce paper experiments

Requirements

• Python 3 with the following libraries
  • numpy
  • pandas
  • networkx
  • scipy
  • gurobi
  • sympy
  • matplotlib
• Gurobi Optimizer 9.1.2

Datasets

Follow the instructions for cricca to generate input files, and store them in the following directories.

• data/pre_preprocessing/lv
• data/pre_preprocessing/tf

Experiments

1. Preprocessing

• In this directory run the following command:
  • python3 src/exp_preprocess.py --first_seed 0 --last_seed 19
• The directory data/post_preprocessing will be created.

2. Kernelization

• In this directory run the following command:
  • python3 src/exp_runkernel.py --first_seed 0 --last_seed 9 --v2
  • Run python3 src/exp_runkernel.py --help for more options.
• The directory data/post_kernels_v2 will be created.

3. Decomposition

• In this directory run the following command:
  • python3 src/exp_runbsd.py --first_seed 0 --last_seed 0 --lp-only --kernel-v2 --perf-lp-v2
  • Run python3 src/exp_runbsd.py --help for more options.
• The directory data/finaldata_kv2_pv2 will be created.
• Output files can be parsed by src/print_pkl.py.