Basic introduction to graphs, neural networks and graph neural networks.ipynb is a Jupyter-notebook of python code snippets reachly intertwined with commentary, that:
- showcases graph manipulation and plotting with
networkxandpython-igraph, - uses
GEMpackage for graph-embeddings, - introduces basics of
pytorchfor defining a (graph) neural network, - follows the discussion in http://tkipf.github.io/graph-convolutional-networks/ to use a neural network to propagate labels through the canonical Zachary's karate network,
- lists useful resources for further development and study.
The notebook was exported as an html for ease of viewing (locally) but You are encouraged to run the notebook Yourself.
You can find versions of the used python packages in the environment.yml file generated by conda.
To use the node2vec embedding from the GEM package, I've downloaded the SNAP library from here without much hassle.
In order to have the python-igraph package not complain during plotting, I've installed it via conda install -c conda-forge python-igraph.