This paper presents a new classification framework for both first and second order statistics, i.e. mean/location and covariance matrix. In the last decade, several covariance matrix classification algorithms have been proposed. They often leverage the Riemannian geometry of symmetric positive definite matrices (SPD) with its affine invariant metric and have shown strong performance in many applications. However, their underlying statistical model assumes a zero mean hypothesis. In practice, it is often estimated and then removed in a preprocessing step. This is of course damaging for applications where the mean is a discriminative feature. Unfortunately, the distance associated to the affine invariant metric for both mean and covariance matrix remains unknown. Leveraging previous works on geodesic triangles, we propose two affine invariant divergences that use both statistics. Then, we derive an algorithm to compute the associated Riemannian centers of mass. Finally, a divergence based Nearest centroid, applied on the crop classification dataset Breizhcrops, shows the interest of the proposed framework.

Neuroscience studies face challenges in gathering large datasets, which limits the use of machine learning (ML) approaches. One possible solution is to incorporate additional data from large public datasets; however, data collected in different contexts often exhibit systematic differences called dataset shifts. Various factors, e.g., site, device type, experimental protocol, or social characteristics, can lead to substantial divergence of brain signals that can hinder the success of ML across datasets. In this work, we focus on dataset shifts in recordings of brain activity using MEG and EEG. State-of-the-art predictive approaches on M/EEG signals classically represent the data by covariance matrices. Model-based dataset alignment methods can leverage the geometry of covariance matrices, leading to three steps: re-centering, re-scaling, and rotation correction. This work explains theoretically how differences in brain activity, anatomy, or device configuration lead to certain shifts in data covariances. Using controlled simulations, the different alignment methods are evaluated. Their practical relevance is evaluated for brain age prediction on one MEG dataset (Cam-CAN, n=646) and two EEG datasets (TUAB, n=1385; LEMON, n=213). When the target sample included recordings from the same subjects with a different task among the same dataset, paired rotation correction was essential (δR2= +0.13 (rest-passive) or +0.17 (rest-smt)). When the target dataset included new subjects and a new task, re-centering led to improved performance (δR2= +0.096 for rest-passive, δR2= +0.045 for rest-smt). For generalization to an independent dataset sampled from a different population and recorded with a different device, re-centering was necessary to achieve brain age prediction performance close to within domain prediction performance. This study demonstrates that the generalization of M/EEG-based regression models across datasets can be substantially enhanced by applying domain adaptation procedures that can statistically harmonize diverse datasets.

This paper studies a statistical model for heteroscedastic (i.e., power fluctuating) signals embedded in white Gaussian noise. Using the Riemannian geometry theory, we propose an unified approach to tackle several problems related to this model. The first axis of contribution concerns parameters (signal subspace and power factors) estimation, for which we derive intrinsic Cramér-Rao bounds and propose a flexible Riemannian optimization algorithmic framework in order to compute the maximum likelihood estimator (as well as other cost functions involving the parameters). Interestingly, the obtained bounds are in closed forms and interpretable in terms of problem's dimensions and SNR. The second axis of contribution concerns the problem of clustering data assuming a mixture of heteroscedastic signals model, for which we generalize the Euclidean K-means++ to the considered Riemannian parameter space. We propose an application of the resulting clustering algorithm on the Indian Pines segmentation problem benchmark.

This paper derives a new change detector for multivariate Synthetic Aperture Radar image time series. Classical statistical change detection methodologies based on covariance matrix analysis are usually built upon the Gaussian assumption, as well as an unstructured signal model. Both of these hypotheses may be inaccurate for high-dimension/resolution images, where the noise can be heterogeneous (non-Gaussian) and where the relevant signals usually lie in a low dimensional subspace (lowrank structure). These two issues are tackled by proposing a new Generalized Likelihood Ratio Test based on a robust (compound Gaussian) low-rank (structured covariance matrix) model. The interest of the proposed detector is assessed on two Synthetic Aperture Radar Image Time Series data set from UAVSAR.

No abstract