Topological and Geometrical Deep Learning

This is the main page for the research subgroup on topological and geometrical deep learning. The main objective of our subgroup is to study neural networks by means of topological and geometrical methods. We focus on both: theoretical analysis of neural networks and applications to computer vision problems. We are associated with the Topological Machine Learning group at the Universitat de Barcelona. We coordinate the reading group in theoretical deep learning.

Members:

Projects:

• Topological Data Analysis for Neural Network Analysis.
• Generative models of neural networks.

Associated publications:

• Preprint: Predicting the generalization gap in neural networks using topological data analysis