GraphSage

The goal of this project is to reproduce the paper "Inductive Representation Learning on Large Graphs".

Folder structure

• baselines/ Code for all the baselines.

• experiments/ Code reproducing all the experiments.

  • fig2a/ Code reproducing the figure 2a of the original paper.
  • fig2b/ Code reproducing the figure 2b of the original paper.
  • fig3/ Code reproducing the figure 3 of the original paper.
  • table1/ Code reproducing the table 1 of the original paper.

• graphsage/ Code related to graphsage models and extensions.

  • datasets/ Code related to datasets and transformations.
  • models/ Code related to models.
  • layers/ Code related to reusable pytorch modules.
  • samplers/ Code related to samplers.
  • trainers/ Custom abstractions to monitor training.

• examples/ Direct application of models on datasets.

• docs/ Auto-Generated & manual code documentation.

• scripts/ Contains bash scripts, this scripts might just be launchers for python scripts defined in the main package. Useful for running long experiments for example.

• data/ Auto-generated, contains original or intermediate synthetic data.

• examples/ All the examples, python scripts or notebooks, illustrating the usage of the package.

• graphsage/ Python package containing the main code for this project.

• results/ Auto-generated, For results, e.g. tables (csv files), and plots (images)

Installation

# Create the conda environment
conda env create -f env.yml
# Add the environment to your jupyter kernels 
python -m ipykernel install --user --name graphsage 
# Activate the environment
conda activate graphsage

Tests Documentation

To generate documentation, run

scripts/makedoc.sh

The documentation entrypoint will be generated at docs/_build/html/index.html