A statistical interpolation code for ocean analysis and forecasting

We present a data assimilation package for use with ocean circulation models in analysis, forecasting and system evaluation applications. The basic functionality of the package is centered on a multivariate linear statistical estimation for a given predicted/background ocean state, observations and error statistics. Novel features of the package include support for multiple covariance models, and the solution of the least squares normal equations either using the covariance matrix or its inverse - the information matrix. The main focus of this paper, however, is on the solution of the analysis equations using the information matrix, which offers several advantages for solving large problems efficiently. Details of the parameterization of the inverse covariance using Markov Random Fields are provided and its relationship to finite difference discretizations of diffusion equations are pointed out. The package can assimilate a variety of observation types from both remote sensing and in-situ platforms. The performance of the data assimilation methodology implemented in the package is demonstrated with a yearlong global ocean hindcast with a 1/4°ocean model. The code is implemented in modern Fortran, supports distributed memory, shared memory, multi-core architectures and uses Climate and Forecasts compliant Network Common Data Format for Input/Output. The package is freely available with an open source license from www.tendral.com/tsis/

Graph-based prior and forward models for inverse problems on manifolds with boundaries

This paper develops manifold learning techniques for the numerical solution of PDE-constrained Bayesian inverse problems on manifolds with boundaries. We introduce graphical Matérn-type Gaussian field priors that enable flexible modeling near the boundaries, representing boundary values by superposition of harmonic functions with appropriate Dirichlet boundary conditions. We also investigate the graph-based approximation of forward models from PDE parameters to observed quantities. In the construction of graph-based prior and forward models, we leverage the ghost point diffusion map algorithm to approximate second-order elliptic operators with classical boundary conditions. Numerical results validate our graph-based approach and demonstrate the need to design prior covariance models that account for boundary conditions.

Robust Acoustic Scene Classification in thePresence of Active Foreground Speech

We present an iVector based Acoustic Scene Clas-sification (ASC) system suited for real life settings where activeforeground speech can be present. In the proposed system, eachrecording is represented by a fixed-length iVector that modelsthe recording’s important properties. A regularized Gaussianbackend classifier with class-specific covariance models is usedto extract the relevant acoustic scene information from theseiVectors. To alleviate the large performance degradation when aforeground speaker dominates the captured signal, we investigatethe use of the iVector framework on Mel-Frequency CepstralCoefficients (MFCCs) that are derived from an estimate of thenoise power spectral density. This noise-floor can be extracted in astatistical manner for single channel recordings. We show that theuse of noise-floor features is complementary to multi-conditiontraining in which foreground speech is added to training signalto reduce the mismatch between training and testing conditions.Experimental results on the DCASE 2016 Task 1 dataset showthat the noise-floor based features and multi-condition trainingrealize significant classification accuracy gains of up to more than25 percentage points (absolute) in the most adverse conditions.These promising results can further facilitate the integration ofASC in resource-constrained devices such as hearables.

Testing covariance models for MEG source reconstruction of hippocampal activity

Beamforming is one of the most commonly used source reconstruction methods for magneto- and electroencephalography (M/EEG). One underlying assumption, however, is that distant sources are uncorrelated and here we tested whether this is an appropriate model for the human hippocampal data. We revised the Empirical Bayesian Beamfomer (EBB) to accommodate specific a-priori correlated source models. We showed in simulation that we could use model evidence (as approximated by Free Energy) to distinguish between different correlated and uncorrelated source scenarios. Using group MEG data in which the participants performed a hippocampal-dependent task, we explored the possibility that the hippocampus or the cortex or both were correlated in their activity across hemispheres. We found that incorporating a correlated hippocampal source model significantly improved model evidence. Our findings help to explain why, up until now, the majority of MEG-reported hippocampal activity (typically making use of beamformers) has been estimated as unilateral.

Accuracy of Components of the SCAT5 and ChildSCAT5 to Identify Children with Concussion

The Sport Concussion Assessment Tool 5th Edition (SCAT5) is a standardized measure of concussion. In this prospective observational study, the ability of the SCAT5 and ChildSCAT5 to differentiate between children with and without a concussion was examined. Concussed children (n=91) and controls (n=106) were recruited from an emergency department in three equal-sized age bands (5–8/9–12/13–16 years). Analysis of covariance models (adjusting for participant age) were used to analyze group differences on components of the SCAT5. On the SCAT5 and ChildSCAT5, respectively, youth with concussion reported a greater number (d=1.47; d=0.52) and severity (d=1.27; d=0.72) of symptoms than controls (all p<0.001). ChildSCAT5 parent-rated number (d=0.98) and severity (d=1.04) of symptoms were greater for the concussion group (all p<0.001). Acceptable levels of between-group discrimination were identified for SCAT5 symptom number (AUC=0.86) and severity (AUC=0.84) and ChildSCAT5 parent-rated symptom number (AUC=0.76) and severity (AUC=0.78). Our findings support the utility of the SCAT5 and ChildSCAT5 to accurately distinguish between children with and without a concussion.

On the Family of Covariance Functions Based on ARMA Models

In time series analyses, covariance modeling is an essential part of stochastic methods such as prediction or filtering. For practical use, general families of covariance functions with large flexibilities are necessary to model complex correlations structures such as negative correlations. Thus, families of covariance functions should be as versatile as possible by including a high variety of basis functions. Another drawback of some common covariance models is that they can be parameterized in a way such that they do not allow all parameters to vary. In this work, we elaborate on the affiliation of several established covariance functions such as exponential, Matérn-type, and damped oscillating functions to the general class of covariance functions defined by autoregressive moving average (ARMA) processes. Furthermore, we present advanced limit cases that also belong to this class and enable a higher variability of the shape parameters and, consequently, the representable covariance functions. For prediction tasks in applications with spatial data, the covariance function must be positive semi-definite in the respective domain. We provide conditions for the shape parameters that need to be fulfilled for positive semi-definiteness of the covariance function in higher input dimensions.

Testing covariance models for MEG source reconstruction of hippocampal activity

Beamforming is one of the most commonly used source reconstruction methods for magneto- and electroencephalography (M/EEG). One underlying assumption, however, is that distant sources are uncorrelated and here we tested whether this is an appropriate model for the human hippocampal data. We revised the Empirical Bayesian Beamfomer (EBB) to accommodate specific a-priori correlated source models. We showed in simulation that we could use model evidence (as approximated by Free Energy) to distinguish between different correlated and uncorrelated source scenarios. Using group MEG data in which the participants performed a hippocampal-dependent task, we explored the possibility that the hippocampus or the cortex or both were correlated in their activity across hemispheres. We found that incorporating a correlated hippocampal source model significantly improved model evidence. Our findings help to explain why, up until now, the majority of MEG-reported hippocampal activity (typically making use of beamformers) has been estimated as unilateral.