yung1231 / Implementation-GNN

PyTorch implementation GCN and DGCNN paper

Implementation of GNN

PyTorch implementation of Graph Convolutional Networks (GCN) and Deep Graph CNN (DGCNN)

Reference Papers

GCN is based on the paper presented in ICLR 2017:

• Semi-supervised classification with graph convolutional networks

DGCNN is based on the paper presented in AAAI 2018:

• An End-to-End Deep Learning Architecture for Graph Classification

Requirements

• PyTorch
• PyTorch Geometric

Datasets

This implementation uses the Lab dataset which includes two classes:

• Benign
• Mirai

Results

The results of the implementation are shown in the following table:

Detector	Training Accuracy	Testing Accuracy
GCN	99.43%	99.36%
DGCN	99.56%	99.53%