This repository contains the code used to produce the results for the bacheor thesis "Classification of Noise Sources in LIGO Gravitational Wave Observations using DeepNeural Networks and Model Explainability Analysis". For detailed information about the project and results, please read the thesis.
The file "model.py" contains the CNN used for classification, as well as a confusion matrix and a list of the most confused instances by the model.
The file "post_hoc_analysis.py" contains the analysis of the trained model using various saliency and class activation maps.