Timezone: »
Covariance matrices have attracted attention for machine learning applications due to their capacity to capture interesting structure in the data. The main challenge is that one needs to take into account the particular geometry of the Riemannian manifold of symmetric positive definite (SPD) matrices they belong to. In the con- text of deep networks, several architectures for these matrices have recently been proposed. In our article, we introduce a Riemannian batch normalization (batch- norm) algorithm, which generalizes the one used in Euclidean nets. This novel layer makes use of geometric operations on the manifold, notably the Riemannian barycenter, parallel transport and non-linear structured matrix transformations. We derive a new manifold-constrained gradient descent algorithm working in the space of SPD matrices, allowing to learn the batchnorm layer. We validate our proposed approach with experiments in three different contexts on diverse data types: a drone recognition dataset from radar observations, and on emotion and action recognition datasets from video and motion capture data. Experiments show that the Riemannian batchnorm systematically gives better classification performance compared with leading methods and a remarkable robustness to lack of data.
Author Information
Daniel Brooks (Thales - LIP6)
Olivier Schwander (Sorbonne Université)
Frederic Barbaresco (THALES LAND & AIR SYSTEMS)
Jean-Yves Schneider (THALES LAND & AIR SYSTEMS)
Matthieu Cord (Sorbonne University)
More from the Same Authors
-
2020 : Paper 16: Driving Behavior Explanation with Multi-level Fusion »
Matthieu Cord · Patrick Pérez -
2021 Poster: RED : Looking for Redundancies for Data-FreeStructured Compression of Deep Neural Networks »
Edouard YVINEC · Arnaud Dapogny · Matthieu Cord · Kevin Bailly -
2021 Poster: Look at the Variance! Efficient Black-box Explanations with Sobol-based Sensitivity Analysis »
Thomas FEL · Remi Cadene · Mathieu Chalvidal · Matthieu Cord · David Vigouroux · Thomas Serre -
2019 Poster: RUBi: Reducing Unimodal Biases for Visual Question Answering »
Remi Cadene · Corentin Dancette · Hedi Ben younes · Matthieu Cord · Devi Parikh -
2019 Poster: Zero-Shot Semantic Segmentation »
Maxime Bucher · Tuan-Hung VU · Matthieu Cord · Patrick Pérez -
2019 Poster: Addressing Failure Prediction by Learning Model Confidence »
Charles Corbière · Nicolas THOME · Avner Bar-Hen · Matthieu Cord · Patrick Pérez -
2018 Poster: Revisiting Multi-Task Learning with ROCK: a Deep Residual Auxiliary Block for Visual Detection »
Taylor Mordan · Nicolas THOME · Gilles Henaff · Matthieu Cord -
2013 Poster: Top-Down Regularization of Deep Belief Networks »
Hanlin Goh · Nicolas Thome · Matthieu Cord · Joo-Hwee Lim