Experimental Pipeline for Renormalized Graph Representations for Node Classification

This repository contains the code to reproduce the experimental pipeline described in our paper.

File Overview

• SpecificHeat.py and SpecificHeat.ipynb
  Calculate the characteristic scales using spectral entropy values.

• DataGeneratorCharacteristic.py
  Preprocess graphs by aligning them to the characteristic scales.

• new_train.py
  Run experiments on the preprocessed graphs.
  (You can also refer to run_cora_all_t_all_epochs as an example.)

• statistics_and_graph_generator_single_node.py
  Generate metrics, perform statistical tests, and produce the graphs discussed in the paper.

Experimental Pipeline

The steps above complete the full experimental pipeline described in the paper.

Visualization of the Rewiring Procedure

Paper

The associated paper is available on arXiv.

License

This project is licensed under the MIT License.