S&P: Towards Scalable and Powerful Graph Learning with Hierarchical Structural Acquisition

This is the official implementation of the paper "S&P: Towards Scalable and Powerful Graph Learning with Hierarchical Structural Acquisition".

---

Quickstart

1. Configure hyper-parameters

   Edit config.yaml to set dataset names, model structure, and training options.

2. Preprocess data

   Generate structural features and serialized .pt artifacts:

   python preprocess.py --dataset <dataset_name>

   Outputs are saved to processed/<dataset>_*.pt based on the configuration flags.

3. Train and evaluate

   python main.py --dataset <dataset_name>