Python3 implemetation of DeepGL.

• This package is implemented for: Fujiwara et al., Network Comparison with Interpretable Contrastive Network Representation Learning, JDSSV, 2022.
• Related repository: Contrastive Network Representation Learning (cNRL), https://github.com/takanori-fujiwara/cnrl
• Original DeepGL paper: Rossi et al., Deep Inductive Graph Representation Learning, IEEE TKDE, 2018.

Current implementation supports a major portion of DeepGL. However, for example, local graphlet count-based features are not supported. These functionality will be tentatively implemented in the future.

• Note: Tested on macOS Sonoma and Ubuntu 20.0.4 LTS.
• Windows is not supported because graph-tool is not available for Windows.

Install graph-tool (https://graph-tool.skewed.de/installation.html)

• For example, macOS with Homebrew (when not using virtual environment),

  brew install graph-tool

• When using virtual environment, there are two options:

  • Option 1. Follow the graph-tool instruction (need a lot of time for compiling graph-tool).

    • Check a section of "Installing in a virtualenv".

    • graph-tool's instruction doesn't support Python3.12. For Python3.12, before the configure step (i.e., ./configure --prefix=$HOME/.local), run commands below:

      pip3 install setuptools pycairo

  • Option 2. Use virtual environment with "include-system-site-packages = true"

    • Either make venv with --system-site-packages option (e.g., python3 -m venv --system-site-packages venv) or edit "pyenv.cfg" (include-system-site-packages = true)

    • then

      brew install graph-tool

Install with pip3. Move to the directory of this repository. Then,

pip3 install .