Search

The time-evolving precision matrix of a piecewise-constant Gaussian graphical model encodes the dynamic conditional dependency structure of a multivariate time-series. Traditionally, graphical models are estimated under the assumption that data is drawn identically from a generating distribution. Introducing sparsity and sparse-difference inducing priors we relax these assumptions and propose a novel regularized M-estimator to jointly estimate both the graph and changepoint structure. The resulting estimator possesses the ability to therefore favor sparse dependency structures and/or smoothly evolving graph structures, as required. Moreover, our approach extends current methods to allow estimation of changepoints that are grouped across multiple dependencies in a system. An efficient algorithm for estimating structure is proposed. We study the empirical recovery properties in a synthetic setting. The qualitative effect of grouped changepoint estimation is then demonstrated by applying the method on two real-world data-sets., Comment: 32 pages, 9 figures

Summary Animal models and archived human biobank tissues are useful resources for research in disease development, diagnostics and therapeutics. For the preservation of microscopic anatomical features and to facilitate long‐term storage, a majority of tissue samples are denatured by the chemical treatments required for fixation, paraffin embedding and subsequent deparaffinization. These aggressive chemical processes are thought to modify the biochemical composition of the sample and potentially compromise reliable spectroscopic examination useful for the diagnosis or biomarking. As a result, spectroscopy is often conducted on fresh/frozen samples. In this study, we provide an extensive characterization of the biochemical signals remaining in processed samples (formalin fixation and paraffin embedding, FFPE) and especially those originating from the anatomical layers of a healthy rat colon. The application of chemometric analytical methods (unsupervised and supervised) was shown to eliminate the need for tissue staining and easily revealed microscopic features consistent with goblet cells and the dense populations of cells within the mucosa, principally via strong nucleic acid signals. We were also able to identify the collagenous submucosa‐ and serosa‐ as well as the muscle‐associated signals from the muscular regions and blood vessels. Applying linear regression analysis to the data, we were able to corroborate this initial assignment of cell and tissue types by confirming the biological origin of each layer by reference to a subset of authentic biomolecular standards. Our results demonstrate the potential of using label‐free Raman microspectroscopy to obtain superior imaging contrast in FFPE sections when compared directly to conventional haematoxylin and eosin (H&E) staining.

Wavelet energy-based lacunarity features, which measure deviations from translational statistical invariance over multiple scales, were recently proposed for object detection and classification in sonar imagery. The authors here extend the idea to incorporate further robustness to background type whilst retaining sensitivity to local changes in texture caused by the presence of man-made objects. The resulting textural-lacunarity features are constructed by estimating the joint distribution of local neighbourhoods with empirical distributions over an adaptive texton dictionary. Experiments on a synthetic aperture sonar imagery dataset suggest that the features offer significant improvements in the receiver operating curve.

International audience; Recent work has shown that existing powerful Bayesian hyperspectral unmixing algorithms can be significantly improved by incorporating the inherent local spatial correlations between pixel class labels via the use of Markov random fields. We here propose a new Bayesian approach to joint hyperspectral unmixing and image classification such that the previous assumption of stochastic abundance vectors is relaxed to a formulation whereby a common abundance vector is assumed for pixels in each class. This allows us to avoid stochastic reparameterizations and, instead, we propose a symmetric Dirichlet distribution model with adjustable parameters for the common abundance vector of each class. Inference over the proposed model is achieved via a hybrid Gibbs sampler, and in particular, simulated annealing is introduced for the label estimation in order to avoid the local- trap problem. Experiments on a synthetic image and a popular, publicly available real data set indicate the proposed model is faster than and outperforms the existing approach quantitatively and qualitatively. Moreover, for appropriate choices of the Dirichlet parameter, it is shown that the proposed approach has the capability to induce sparsity in the inferred abundance vectors. It is demonstrated that this offers increased robustness in cases where the preprocessing endmember extraction algorithms overestimate the number of active endmembers present in a given scene.

Hyperspectral unmixing methods often use a conventional least squares based lasso which assumes that the data follows the Gaussian distribution. The normality assumption is an approximation which is generally invalid for real imagery data. We consider a robust (non-Gaussian) approach to sparse spectral unmixing of remotely sensed imagery which reduces the sensitivity of the estimator to outliers and relaxes the linearity assumption. The method consists of several appropriate penalties. We propose to use an ep norm with 0 < p < 1 in the sparse regression problem, which induces more sparsity in the results, but makes the problem non-convex. On the other hand, the problem, though non-convex, can be solved quite straightforwardly with an extensible algorithm based on iteratively reweighted least squares. To deal with the huge size of modern spectral libraries we introduce a library reduction step, similar to the multiple signal classification (MUSIC) array processing algorithm, which not only speeds up unmixing but also yields superior results. In the hyperspectral setting we extend the traditional least squares method to the robust heavy-tailed case and propose a generalised M-lasso solution. M-estimation replaces the Gaussian likelihood with a fixed function ρ(e) that restrains outliers. The M-estimate function reduces the effect of errors with large amplitudes or even assigns the outliers zero weights. Our experimental results on real hyperspectral data show that noise with large amplitudes (outliers) often exists in the data. This ability to mitigate the influence of such outliers can therefore offer greater robustness. Qualitative hyperspectral unmixing results on real hyperspectral image data corroborate the efficacy of the proposed method.

SAR roughness images acquired by satellites are affected by speckle noise. We propose a denoising method based on semi-local estimation of the Hurst exponent in the wavelet domain. To estimate the Hurst parameter, we consider low resolution coefficients to be relatively free of noise. The high resolution level coefficients are most affected by noise and their energy is corrected according to the estimated parameter. The algorithm is tested on ESA Sentinel-1 data single look images. In our experiments, we used the discrete wavelet transform and the dual-tree complex wavelet transform. The denoising system proposed is anisotropic and realizes a performant treatment of the homogeneous regions; the degree of oversmoothing can be adjusted by the number of decomposition levels used in the denoising process.

We investigate where and how key dependency structure between measures of network activity change throughout the course of daily activity. Our approach to data-mining is probabilistic in nature, we formulate the identification of dependency patterns as a regularised statistical estimation problem. The resulting model can be interpreted as a set of time-varying graphs and provides a useful visual interpretation of network activity. We believe this is the first application of dynamic graphical modelling to network traffic of this kind. Investigations are performed on 9 days of real-world network traffic across a subset of IP's. We demonstrate that dependency between features may change across time and discuss how these change at an intra and inter-day level. Such variation in feature dependency may have important consequences for the design and implementation of probabilistic intrusion detection systems.

Atopic dermatitis (AD), a chronic, relapsing, inflammatory skin disease that is characterized by intense pruritus and eczematous lesions with up to 90% of patients presenting with mild to moderate disease. Current topical treatments for AD have not changed in over 15 years and are associated with safety concerns. In AD, overactivity of phosphodiesterase 4 (PDE4), leads to inflammation and disease exacerbation. Crisaborole Topical Ointment, 2%, is a novel, nonsteroidal, topical anti-inflammatory PDE4 inhibitor currently being investigated for the treatment of mild to moderate AD. Preliminary studies in children and adults demonstrated favorable efficacy and safety profiles. Crisaborole may represent an anti-inflammatory option that safely minimizes the symptoms and severity of AD and that can be used for both acute and long-term management.

We here establish and exploit the result that 2D isotropic self-similar fields beget quasi-decorrelated wavelet coefficients and that the resulting localised log sample second moment statistic is asymptotically normal. This leads to the development of a semi-local scaling exponent estimation framework with optimally modified weights. Furthermore, recent interest in penalty methods for least square problems and generalized Lasso for scaling exponent estimation inspires the simultaneous incorporation of both bounding box constraints and total variation smoothing into an iteratively reweighted least-square estimator framework. Numerical results on fractional Brownian fields with global and piecewise constant, semi-local Hurst parameters illustrate the benefits of the new estimators.

A construction paradigm is proposed to refine a class of wavelet bases such that the filter characteristics are enhanced. In particular, it is shown that the entire Mth order B-wavelet family belong to this class. Exploiting some recent work, a fast integral wavelet transform is found for the refined B-wavelet family. Moreover, this refined basis can be placed into the setting of a multiresolution analysis that has a multiplicity greater than one. Accordingly, a fast, discrete, pyramidal algorithm is realised.

The Mobius function is utilised to propose a class of multi-channel, multi-sampling rate theorems. Our construction facilitates a parallel decomposition of the Fourier transform into an array of general transform operators. A practical example is discussed that employs a complex square wave biorthonormal basis.