In this repository we present an implementation of Graph Neural Networks - GNNs for astroparticle application. In particular we will show how is possible using this new paradigm of Neural Networks to process informations which don't present an euclidean structure, so for that data in which convolution or other conventional techniques are not suitable. To implent GNN we use Spektral which is a Python library for graph deep learning, based on the Keras API and TensorFlow 2.
A graph is defined as G = (V,E,A) where V is a set of nodes (or vertex), E is a set of edges (or links) and A is the adjacency matrix. An example of graph is shown in the figure below
A graph can be directed if the edges have a specific direction or undirected if edges don't have any direction. A graph can be weighted if in the edges is given a numerical weight or unweighted if each weight in the link is 1 and equal for all the links.
GNN is a class of artificial neural network for processing data that can be represented as graphs.
https://bulldogjob.com/readme/how-to-write-a-good-readme-for-your-github-project
https://www.datacamp.com/tutorial/comprehensive-introduction-graph-neural-networks-gnns-tutorial
https://www.analyticsvidhya.com/blog/2022/03/what-are-graph-neural-networks-and-how-do-they-work/
To run code it's essential that you've already installed Keras and TensorFlow. To install Spektral the best way is from PyPi
pip install spektralOther libraries present in the code are
pip install pandas, statistics, numpy, sklearn, seaborn, matplotlibTo visualize graph we use the library Netowrkx
pip install networkx