GraphGallery

GraphGallery is a gallery of state-of-the-arts graph neural networks for TensorFlow 2.x.

This repo aims to achieve 4 goals:

• Similar or higher performance
• Faster training and testing
• Simple and convenient to use, high scalability
• Easy to read source codes

Installation

pip install -U graphgallery

Implementations

In detail, the following methods are currently implemented:

Semi-supervised models

General

• ChebyNet from Michaël Defferrard et al, 📝Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering, NIPS'16. [:octocat:Official Codes], [🌈 GraphGallery Example]
• GCN from Thomas N. Kipf et al, 📝Semi-Supervised Classification with Graph Convolutional Networks, ICLR'17. [:octocat:Official Codes], [🌈 GraphGallery Example]
• GraphSAGE from William L. Hamilton et al, 📝Inductive Representation Learning on Large Graphs, NIPS'17. [:octocat:Official Codes], [🌈 GraphGallery Example]
• FastGCN from Jie Chen et al, FastGCN: Fast Learning with Graph Convolutional Networks via Importance Sampling, ICLR'18. [:octocat:Official Codes], [🌈 GraphGallery Example]
• LGCN from Hongyang Gao et al, 📝Large-Scale Learnable Graph Convolutional Networks, KDD'18. [:octocat:Official Codes], [🌈 GraphGallery Example]
• GAT from Petar Veličković et al, 📝Graph Attention Networks, ICLR'18.
  [:octocat:Official Codes], [🌈 GraphGallery Example]
• SGC from Felix Wu et al, 📝Simplifying Graph Convolutional Networks, ICML'19. [:octocat:Official Codes], [🌈 GraphGallery Example]
• GWNN from Bingbing Xu et al, 📝Graph Wavelet Neural Network, ICLR'19. [:octocat:Official Codes], [🌈 GraphGallery Example]
• GMNN from Meng Qu et al, 📝Graph Markov Neural Networks, ICML'19. [:octocat:Official Codes], [🌈 GraphGallery Example]
• ClusterGCN from Wei-Lin Chiang et al, 📝Cluster-GCN: An Efficient Algorithm for Training Deep and Large Graph Convolutional Networks, KDD'19. [:octocat:Official Codes], [🌈 GraphGallery Example]
• DAGNN from Meng Liu et al, 📝Towards Deeper Graph Neural Networks, KDD'20. [:octocat:Official Codes], [🌈 GraphGallery Example]

Defense models

• RobustGCN from Dingyuan Zhu et al, 📝Robust Graph Convolutional Networks Against Adversarial Attacks, KDD'19. [:octocat:Official Codes], [🌈 GraphGallery Example]
• SBVAT/OBVAT from Zhijie Deng et al, 📝Batch Virtual Adversarial Training for Graph Convolutional Networks, ICML'19. [:octocat:Official Codes]

Unsupervised models

• Deepwalk from Bryan Perozzi et al, 📝DeepWalk: Online Learning of Social Representations, KDD'14. [:octocat:Official Codes], [🌈 GraphGallery Example]
• Node2vec from Aditya Grover et al, 📝node2vec: Scalable Feature Learning for Networks, KDD'16. [:octocat:Official Codes], [🌈 GraphGallery Example]

Quick Start

Datasets

from graphgallery.data import Planetoid
# set `verbose=False` to avoid these printed tables
data = Planetoid('cora', verbose=False)
adj, x, labels = data.graph.unpack()
idx_train, idx_val, idx_test = data.split()
# adj:  adjacency matrix: 2D Scipy sparse matrix
# x:  feature matrix: 2D Numpy array
# labels:  class labels: 1D Numpy array
# idx_train:  training indices: 1D Numpy array
# idx_val:  validation indices: 1D Numpy array
# idx_test:  testing indices: 1D Numpy array

currently the supported datasets are:

>>> data.supported_datasets
('citeseer', 'cora', 'pubmed')

Example of GCN model

from graphgallery.nn.models import GCN
# adj is scipy sparse matrix, x is numpy array matrix
model = GCN(adj, x, labels, device='GPU', norm_x='l1', seed=123)
# build your GCN model with default hyper-parameters
model.build()
# train your model. here idx_train and idx_val are numpy arrays
his = model.train(idx_train, idx_val, verbose=1, epochs=100)
# test your model
loss, accuracy = model.test(idx_test)
print(f'Test loss {loss:.5}, Test accuracy {accuracy:.2%}')

On Cora dataset:

<Loss = 1.0161 Acc = 0.9500 Val_Loss = 1.4101 Val_Acc = 0.7740 >: 100%|██████████| 100/100 [00:01<00:00, 118.02it/s]
Test loss 1.4123, Test accuracy 81.20%

Customization

• Build your model you can use the following statement to build your model

# one hidden layer with hidden units 32 and activation function RELU
>>> model.build(hiddens=32, activations='relu')

# two hidden layer with hidden units 32, 64 and all activation functions are RELU
>>> model.build(hiddens=[32, 64], activations='relu')

# two hidden layer with hidden units 32, 64 and activation functions RELU and ELU
>>> model.build(hiddens=[32, 64], activations=['relu', 'elu'])

# other parameters like `dropouts` and `l2_norms` (if have) are the SAME.

• Train your model

# train with validation
>>> his = model.train(idx_train, idx_val, verbose=1, epochs=100)
# train without validation
# his = model.train(idx_train, verbose=1, epochs=100)

here his is tensorflow Histoory like instance (or itself).

• Test you model

loss, accuracy = model.test(idx_test)
print(f'Test loss {loss:.5}, Test accuracy {accuracy:.2%}')

• Display hyper-parameters

You can simply use model.show() to show all your Hyper-parameters. Otherwise you can also use model.show('model') or model.show('train') to show your model parameters and training parameters.

NOTE: you should install texttable first.

Visualization

NOTE: you must install SciencePlots package for a better preview.

• Accuracy

import matplotlib.pyplot as plt
with plt.style.context(['science', 'no-latex']):
    plt.plot(his.history['acc'])
    plt.plot(his.history['val_acc'])
    plt.legend(['Train Accuracy', 'Val Accuracy'])
    plt.ylabel('Accuracy')
    plt.xlabel('Epochs')
    plt.autoscale(tight=True)
    plt.show()    

• Loss

import matplotlib.pyplot as plt
with plt.style.context(['science', 'no-latex']):
    plt.plot(his.history['loss'])
    plt.plot(his.history['val_loss'])
    plt.legend(['Train Loss', 'Val Loss'])
    plt.ylabel('Loss')
    plt.xlabel('Epochs')
    plt.autoscale(tight=True)
    plt.show()    

More Examples

Please refer to the examples directory.

TODO Lists

• Add Docstrings and Documentation
• Add PyTorch models support
• Support for graph Classification and link prediction tasks
• Support for Heterogeneous graphs

Acknowledgement

This project is motivated by Pytorch Geometric, Tensorflow Geometric and Stellargraph, and the original implementations of the authors, thanks for their excellent works!