Repository for the paper "Graph Auto-Encoders for Financial Clustering"

Requirements

Python 3.6
torch
torch_geometric

Instructions

This is a simple code to transform a graph (via it's adjacency matrix) from an excel sheet to a PyTorch Geometric graph object. The graph may be weighted, directed or have self loops, if you don't want any of these then simply delete that part of the spreadsheet (i.e. delete all 'feature x' rows). I created this as there seems to be a real lack of literature on creating your own PyTorch Geometric graphs online.

The excel_example.xls file is included to show the necessary excel layout, simply download and edit it to fit your graph. Note for the adjacency matrix of a weighted graph the (i, j) entry is the edge from node i to node j.

If this is found to be helpful in your work consider refrencing/linking the repository.