EvoCut is a Python library that accelerates Mixed-Integer Linear Programming (MILP) by injecting problem-specific cutting planes into the LP relaxation. These cuts reduce the feasible set of the LP relaxation and improve solver efficiency.

python -m venv .venv
source .venv/bin/activate    # Windows: .venv\Scripts\activate

pip install -r requirements.txt

Note: Python 3.9 recommended. Requires a licensed MILP solver (e.g., Gurobi).

In the configs/ directory:

1. Copy each file whose name contains _template.
2. Rename the copy (remove _template).
3. Fill in credentials and hyperparameters as needed.

Use the relevant preprocessing script in data/:

python data/data_prepare<suffix>.py

Replace <suffix> with the correct option for your dataset variant (e.g., rand).

Run EvoCut on a problem instance:

python src/main.py <args>

See all options with:

python src/main.py -h

To check the optimal solution preservation rate of generated cuts:

python experiments/OSP_cuts.py <args>

Evaluate EvoCut on held-out instances:

python experiments/evaluate_cut.py <args>

• Determinism: seeds are configurable in configs or CLI flags.
• Hardware/solver versions may affect runtime but not correctness.
• Minimal dependencies ensure reproducibility across machines.