Geodesically-convex optimization for averaging partially observed covariance matrices - Archive ouverte HAL Access content directly
Conference Papers Year : 2020

Geodesically-convex optimization for averaging partially observed covariance matrices

Abstract

Symmetric positive definite (SPD) matrices permeates numerous scientific disciplines, including machine learning, optimization, and signal processing. Equipped with a Rieman-nian geometry, the space of SPD matrices benefits from compelling properties and its derived Riemannian mean is now the gold standard in some applications, e.g. brain-computer interfaces (BCI). This paper addresses the problem of averaging covariance matrices with missing variables. This situation often occurs with inexpensive or unreliable sensors, or when artifact-suppression techniques remove corrupted sensors leading to rank deficient matrices, hindering the use of the Riemannian geometry in covariance-based approaches. An alternate but questionable method consists in removing the matrices with missing variables , thus reducing the training set size. We address those limitations and propose a new formulation grounded in geodesic convexity. Our approach is evaluated on generated datasets with a controlled number of missing variables and a known baseline, demonstrating the robustness of the proposed estimator. The practical interest of this approach is assessed on real BCI datasets. Our results show that the proposed average is more robust and better suited for classification than classical data imputation methods.
Fichier principal
Vignette du fichier
yger20a.pdf (664.63 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02984423 , version 1 (30-10-2020)

Identifiers

  • HAL Id : hal-02984423 , version 1

Cite

Florian Yger, Sylvain Chevallier, Quentin Barthélemy, Suvrit Sra. Geodesically-convex optimization for averaging partially observed covariance matrices. Asian Conference on Machine Learning (ACML), Nov 2020, Bangkok, Thailand. pp.417 - 432. ⟨hal-02984423⟩
127 View
125 Download

Share

Gmail Facebook X LinkedIn More