Repository files navigation

The code in this repository implements an efficient generalization of thepopular Convolutional Neural Networks (CNNs) to arbitrary graphs, presented inour paper:

Michaël Defferrard, Xavier Bresson, Pierre Vandergheynst,Convolutional NeuralNetworks on Graphs with Fast Localized Spectral Filtering, NeuralInformation Processing Systems (NIPS), 2016.

Additional material:

• NIPS2016 spotlight video, 2016-11-22.
• Deep Learning on Graphs, a lecture for EPFL's master courseANetwork Tour of Data Science, 2016-12-21.
• Deep Learning on Graphs, an invited talk at theDeep Learning onIrregular Domains workshop of BMVC, 2017-09-17.

There is also implementations of the filters used in:

• Joan Bruna, Wojciech Zaremba, Arthur Szlam, Yann LeCun,Spectral Networksand Locally Connected Networks on Graphs, International Conference onLearning Representations (ICLR), 2014.
• Mikael Henaff, Joan Bruna and Yann LeCun,Deep Convolutional Networks onGraph-Structured Data, arXiv, 2015.

1. Clone this repository.

   git clone https://github.com/mdeff/cnn_graphcd cnn_graph

2. Install the dependencies. The code should run with TensorFlow 1.0 and newer.

   pip install -r requirements.txt# or make install

3. Play with the Jupyter notebooks.

   jupyter notebook

Run all the notebooks to reproduce the experiments onMNIST and20NEWS presented inthe paper.

cd nips2016make

To use our graph ConvNet on your data, you need:

1. a data matrix where each row is a sample and each column is a feature,
2. a target vector,
3. optionally, an adjacency matrix which encodes the structure as a graph.

See theusage notebook for a simple example with fabricated data.Please get in touch if you are unsure about applying the model to a differentsetting.

The code in this repository is released under the terms of theMIT license.Please cite ourpaper if you use it.

@inproceedings{cnn_graph,  title = {Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering},  author = {Defferrard, Micha\"el and Bresson, Xavier and Vandergheynst, Pierre},  booktitle = {Advances in Neural Information Processing Systems},  year = {2016},  url = {https://arxiv.org/abs/1606.09375},}