Spektral is a Python library for graph deep learning, based on the Keras API. The main goal of this project is to provide a simple but flexible framework for creating graph neural networks (GNNs).
You can use Spektral for classifying the nodes of a network, predicting molecular properties, generating new graphs with GANs, clustering nodes, predicting links, and any other task where data is described by graphs.
Spektral implements some of the most popular layers for graph deep learning, including:
- Graph convolutional networks (GCN)
- Chebyshev networks (ChebNets)
- GraphSage
- Edge-conditioned convolutions (ECC)
- Graph attention networks (GAT)
- ARMA convolutions
- Approximated personalized propagation of neural predictions (APPNP)
- Graph isomorphism networks (GIN)
You can also find pooling layers (including global readouts and graph coarsening layers), and lots of utilities to apply graph deep learning in your projects.
See how to get started with Spektral and have a look at the examples for some project templates.
The source code of the project is available on Github.
Read the documentation here.
Spektral is compatible with Python 3.5+, and is tested on Ubuntu 16.04 and 18.04. Other Linux distros and MacOS should work as well, but Windows is not supported for now.
To install the required dependencies on Ubuntu run:
$ sudo apt install graphviz libgraphviz-dev libcgraph6Some features of Spektral also require the following optional dependencies:
- RDKit, a library for cheminformatics and molecule manipulation (available through Anaconda);
- dyfunconn, a library to build functional connectivity networks (available through PyPi);
The simplest way to install Spektral is from PyPi:
$ pip install spektralTo install Spektral from source, run this in a terminal:
$ git clone https://github.com/danielegrattarola/spektral.git
$ cd spektral
$ python setup.py install # Or 'pip install .'Note that the setup.py script will not attempt to install a backend for Keras, in order to not mess up any previous installation.
It will, however, install Keras and its dependencies via PyPi (which may include the CPU version of TensorFlow).
If you are already a Keras user, this should not impact you. If you're just getting started, then you may want to install the GPU version of Tensorflow before installing Spektral.
Also note that some features of Spektral may depend explicitly on TensorFlow, although this dependency will be kept to a minimum.
| WARNING |
|---|
| Spektral is still a work in progress and may change significantly before the first release. The API is not mature enough to be considered stable, but we'll try to keep breaking changes to a minimum. |
Spektral is an open source project available on Github, and contributions of all types are welcome. Feel free to open a pull request if you have something interesting that you want to add to the framework.