Towards automatic mining and description of complex deep learning algorithms

Summary

The aim of this thesis is to become a springboard towards the automatic mining and description of complex deep learning algorithms. The implementation targets to create a more abstract view of neural network algorithms. It focuses on the foundational pieces of deep learning (network architecture, training, evaluation) and the recognition of the structure of as many neural network algorithms as possible. Finally, it retains the information retrieved in a database convenient for storing data that are described through an ontology.