Second-order networks in PyTorch

Daniel Brooks (1, 2) , Olivier Schwander (2) , Frédéric Barbaresco (3) , Jean-Yves Schneider , Matthieu Cord (2)

1 Thales LAS France
2 MLIA - Machine Learning and Information Access
3 Thales Air Systems

Classification of Symmetric Positive Definite (SPD) matrices is gaining momentum in a variety machine learning application fields. In this work we propose a Python library which implements neural networks on SPD matrices, based on the popular deep learning framework Pytorch.

Mots clés: SPD matrix; covariance; second-order neural network; Rie- mannian machine learning

Domaines: Intelligence artificielle [cs.AI]

Daniel Brooks, Olivier Schwander, Frédéric Barbaresco, Jean-Yves Schneider, Matthieu Cord. Second-order networks in PyTorch. GSI 2019 - 4th International Conference on Geometric Science of Information, Aug 2019, Toulouse, France. pp.751-758, ⟨10.1007/978-3-030-26980-7_78⟩. ⟨hal-02290841⟩