Covariance and Subspace Inference: Handling Robustness, Variability and Heterogeneity

Explore covariance matrix inference and principal component analysis for non-regular data in heterogeneous environments in this 38-minute talk. Delve into mixed effects models for analyzing repeated measures data in signal processing applications. Learn about classical strategies for Gaussian assumptions and discover an expectation-maximization-based algorithm to handle outliers when the Gaussian assumption fails. Examine a parallel scheme to reduce computational costs and processor overload. Investigate extensions for dealing with missing data in individual responses. Apply these concepts to calibration and imaging in large radio-interferometers. Gain insights from Nabil El Korso of L2S/CentraleSupélec on handling robustness, variability, and challenges in covariance and subspace inference.