Max Flow Interdiction with Graph Neural Networks

Official implementation of KDD2025 paper: Network Interdiction Goes Neural.

🚀 Quick Start

Generate Dataset

# Generate a small dataset for testing (100 instances)
python mfi_generation.py --n_v 20 --budget 10 --num_instances 100

# Generate a larger dataset for training (4000 instances)
python mfi_generation.py --n_v 30 --budget 15 --num_instances 4000

# Custom parameters
python mfi_generation.py --n_v 50 --budget 25 --num_instances 1000 --save_path ./custom_data/

Train the Model

# Basic training
python train.py

# Custom training parameters
python train.py --data_root ./data/maxflow_30_15 --batch_size 64 --num_epochs 200

# Training with optimization test
python train.py --run_optimization

---