dataset comes from this paper plus repo https://github.com/squareRoot3/Rethinking-Anomaly-Detection
The preprocess and save data steps are optional
The preprocess folder contains the requirements.txt file and a jupyter notebook used to transform the original dataformat into csv files.
This assumes the tfinance
dataset to be in the data/raw folder.
This transforms the .csv files into a GNNGraph format and saves it as a .jld2
file.
You can find the result of this in data/processed/
.
You will need https://git-lfs.github.com/ to extract the file from git.
This loads the tfinance.jld2
file and creates scatterplot / gif from the t-SNE-pi reduction.
A GNN model to test with the MNIST for graphs
A first setup, adapting the cora script for tfinance