Authors: Xingyu Chen

A Evoluation Plot for Graph Convolution Network on 4-label dataset

GCN training on the the Karate Club Netowkr Dataset

A Evoluation Plot for Graph Convolution Network on 2-label dataset

GCN training on the the Karate Club Netowkr Dataset

A model depth analysis for Graph Convolution Network. 5-fold cross validation perforamnce of GCN and GCN with residual connections on the 3 Benchmark Datasets. Shaded area indicate the variance of change

• data: contains the Karate club toy dataset, three benmark dataset for semi-supervised learning: Cora, Citeseer and Pubmed, and a large network dataset:GSN

• images: a folder of genrated visualzation plot

• video: two videos of showing how GCN model improves during the traning process (GCN model applied on the toy dataset)

• models.py: It contains the GCN, GCN with residual connection and GAT models

• spectrum_embedding.py: It contains the code to run spectrum embedding model under a supervised task

• tsne.py: It contains the code to run tsne model under a supservised task (Note for t-sne we use the off the shelf implementation from scikit-learn)

• train_utlis: utlis class for training neural model including metrics and evaltion script

• plot_utlis: utlis class for plotting graph

• plot.py: a script to plot results for showing impact of nerual model depth and contain a plot function for wall clock time analysis

• visualization.ipynb : a notebook contains the code to genarate visualizations for qualitative analysis.

• Model.ipynb : a notebook contains the code to compare model perforances quantatively and contains code segment that demnonstartes how to run the the implemented models. The notebook is to analyze semi-supervised perforamnce of the implemented graph nerual model.