Search

We consider probabilistic time-series models for systems that gradually transition among a finite number of states. We are particularly motivated by applications such as human activity analysis where the observed time-series contains segments representing distinct activities such as running or walking as well as segments characterized by continuous transition among these states. Accordingly, the dynamical Wasserstein barycenter (DWB) model introduced in Cheng et al. in 2021 [1] associates with each state, which we call a pure state, its own probability distribution, and models these continuous transitions with the dynamics of the barycentric weights that combine the pure state distributions via the Wasserstein barycenter. Here, focusing on the univariate case where Wasserstein distances and barycenters can be computed in closed form, we extend [1] by discussing two challenges associated with learning a DWB model and two improvements. First, we highlight the issue of uniqueness in identifying the model parameters. Secondly, we discuss the challenge of estimating a dynamically evolving distribution given a limited number of samples. The uncertainty associated with this estimation may cause a model's learned dynamics to not reflect the gradual transitions characteristic of the system. The first improvement introduces a regularization framework that addresses this uncertainty by imposing temporal smoothness on the dynamics of the barycentric weights while leveraging the understanding of the non-uniqueness of the problem. This is done without defining an entire stochastic model for the dynamics of the system as in [1]. Our second improvement lifts the Gaussian assumption on the pure states distributions in [1] by proposing a quantile-based non-parametric representation. We pose model estimation in a variational framework and propose a finite approximation to the infinite dimensional problem.

We pursue tractable Bayesian analysis of generalized linear models (GLMs) for categorical data. Thus far, GLMs are difficult to scale to more than a few dozen categories due to non-conjugacy or strong posterior dependencies when using conjugate auxiliary variable methods. We define a new class of GLMs for categorical data called categorical-from-binary (CB) models. Each CB model has a likelihood that is bounded by the product of binary likelihoods, suggesting a natural posterior approximation. This approximation makes inference straightforward and fast; using well-known auxiliary variables for probit or logistic regression, the product of binary models admits conjugate closed-form variational inference that is embarrassingly parallel across categories and invariant to category ordering. Moreover, an independent binary model simultaneously approximates multiple CB models. Bayesian model averaging over these can improve the quality of the approximation for any given dataset. We show that our approach scales to thousands of categories, outperforming posterior estimation competitors like Automatic Differentiation Variational Inference (ADVI) and No U-Turn Sampling (NUTS) in the time required to achieve fixed prediction quality., Comment: to appear at ICML 2022

Many time series can be modeled as a sequence of segments representing high-level discrete states, such as running and walking in a human activity application. Flexible models should describe the system state and observations in stationary "pure-state" periods as well as transition periods between adjacent segments, such as a gradual slowdown between running and walking. However, most prior work assumes instantaneous transitions between pure discrete states. We propose a dynamical Wasserstein barycentric (DWB) model that estimates the system state over time as well as the data-generating distributions of pure states in an unsupervised manner. Our model assumes each pure state generates data from a multivariate normal distribution, and characterizes transitions between states via displacement-interpolation specified by the Wasserstein barycenter. The system state is represented by a barycentric weight vector which evolves over time via a random walk on the simplex. Parameter learning leverages the natural Riemannian geometry of Gaussian distributions under the Wasserstein distance, which leads to improved convergence speeds. Experiments on several human activity datasets show that our proposed DWB model accurately learns the generating distribution of pure states while improving state estimation for transition periods compared to the commonly used linear interpolation mixture models., To appear at Neurips 2021

Additional file 2: Figures S1-S13 and Table S2. Fig S1 – Phylogenetic tree. Fig S2 – MICs in Canada & the UK. Fig S3 - The effects of candidate determinants on MIC for beta-lactams Fig S4 - The effects of candidate determinants on MIC for MLSB. Fig S5 - The effects of candidate alleles on MIC for tetracyclines. Fig S6 - The effects of candidate alleles on MIC for fluoroquinolone. Fig S7 - The effects of candidate alleles on MIC for the aminoglycoside, spectinomycin. Fig S8 - The effects of candidate alleles on MIC for the pleuromutilin, tiamulin. Fig S9 - The effects of candidate alleles on MIC for trimethoprim (TMP). Fig S10 - Variation in the presence of candidate AMR determinants explains consistent differences between genetic clusters. Fig S11 - Methods of using candidate determinants to predict MIC. Fig S12 - Allelic variation in ermB and unidentified sources of epistasis. Fig S13 -High levels of ‘nestedness’ suggest that resistance determinants to beta-lactams are acquired in a particular order. Table S2 - Binomial tests showing that novel AMR variants are independently associated with MIC in different genetic clusters.

Antimicrobial resistance (AMR) is a major problem for human and animal health, and can co-vary with features of the bacterial genome, its ecological context, and with other traits. Here, we present an integrated study of AMR in the porcine zoonotic pathogen, Streptococcus suis . Collecting a diverse panel of 678 isolates, we obtained replicated measures of Minimum Inhibitory Concentration (MIC) for 16 widely used antimicrobials, from eight distinct classes. Then using whole genome sequences for each isolate, we identified 42 candidate AMR alleles, of which 21 were novel for S. suis . The presence of these alleles explains most of the observed variation in MIC but there was also unexplained variation, partly due to alleles of small effect, and epistatic interactions. We find that MIC varies with S. suis genetic background and country, but less so with clinical phenotype, serotype or over time. These ecological patterns are largely explained by allele prevalence. MIC values for different antibiotic classes positively correlate with one another, but there is limited evidence of direct pleiotropy, where single alleles confer resistance to multiple drug classes. Instead, this covariation is due to alleles for separate classes appearing together, but not necessarily in the same genomic region. The result is a significant excess of multi-drug resistant isolates. Indeed, we found more isolates that carried resistance alleles to five classes of antibiotics than isolates than carried no resistance alleles. We discuss the implications of our results for combatting AMR in S. suis , and for understanding multi-drug resistance in bacteria. Importance Streptococcus suis is a bacterium that causes serious disease with significant welfare and economic impacts for the global pig industry, and meningitis in humans. Antibiotics are the principal treatment for these bacteria, and so the evolution of antibiotic resistance presents a major problem for global food security and human health. We tested susceptibility to 16 different antibiotics in a global sample of 678 S. suis isolates. Using comparative analysis of whole genome sequences, we identified existing and novel genetic variants that affect resistance. Results show that, in most cases, the presence/absence of these alleles is a good predictor of resistance levels, which could help facilitate resistance monitoring through genome sequencing. We also observe a tendency for AMR variants to co-occur together in some S. suis isolates. These multi-drug resistant isolates should be monitored as a priority.

The opportunistic pathogen Streptococcus pneumoniae is commonly carried asymptomatically in the human nasopharynx. Due to high rates of cocolonization with other pneumococcus strains, intraspecific competitive interactions partly determine the carriage duration of strains and thereby their potential to cause disease. These interactions may be mediated by bacteriocins, such as the type IIb bacteriocins encoded by the blp (bacteriocin-like peptide) locus. To understand blp diversity and evolution, we undertook a bioinformatic analysis of 4,418 pneumococcal genomes, including 168 newly sequenced genomes. We describe immense variation at all levels of genomic organization: Gene presence/absence, gene order, and allelic diversity. If we make the extreme and naive hypothesis that assumes all genes in this operon can assort randomly, this variation could lead to 10(15) distinct bacteriocin-related phenotypes, each potentially representing a unique ecological strategy; however, we provide several explanations for why this extreme is not realized. Although rarefaction analysis indicates that the number of unique strategies is not saturated, even after sampling thousands of genomes, we show that the variation is neither unbounded nor random. We delimit three bacteriocin groups, which contain group-specific bacteriocins, immunity genes, and blp operon gene order, and argue that this organization places a constraint on realized ecological strategies. We additionally show that ecological strategy diversity is significantly constrained by pneumococcal phylogeny and clonal structure. By examining patterns of association between alleles within the blp operon, we show that bacteriocin genes, which were believed to function in pairs, can be found with a broad diversity of partner alleles and immunity genes; this overall lack of allelic fidelity likely contributes to the fluid structure of this operon. Our results clarify the diversity of antagonistic ecological strategies in the global pneumococcal population and highlight the potential role of blp bacteriocins in competition within the nasopharynx.

In this paper we demonstrate the utility of fusing energy-resolved observations of Compton scattered photons with traditional attenuation data for the joint recovery of mass density and photoelectric absorption in the context of limited view tomographic imaging applications. We begin with the development of a physical and associated numerical model for the Compton scatter process. Using this model, we propose a variational approach recovering these two material properties. In addition to the typical data-fidelity terms, the optimization functional includes regularization for both the mass density and photoelectric coefficients. We consider a novel edge-preserving method in the case of mass density. To aid in the recovery of the photoelectric information, we draw on our recent method in \cite{r15} and employ a non-local regularization scheme that builds on the fact that mass density is more stably imaged. Simulation results demonstrate clear advantages associated with the use of both scattered photon data and energy resolved information in mapping the two material properties of interest. Specifically, comparing images obtained using only conventional attenuation data with those where we employ only Compton scatter photons and images formed from the combination of the two, shows that taking advantage of both types of data for reconstruction provides far more accurate results.

Multi-task/Multi-output learning seeks to exploit correlation among tasks to enhance performance over learning or solving each task independently. In this paper, we investigate this problem in the context of Gaussian Processes (GPs) and propose a new model which learns a mixture of latent processes by decomposing the covariance matrix into a sum of structured hidden components each of which is controlled by a latent GP over input features and a "weight" over tasks. From this sum structure, we propose a parallelizable parameter learning algorithm with a predetermined initialization for the "weights". We also notice that an ensemble parameter learning approach using mini-batches of training data not only reduces the computation complexity of learning but also improves the regression performance. We evaluate our model on two datasets, the smaller Swiss Jura dataset and another relatively larger ATMS dataset from NOAA. Substantial improvements are observed compared with established alternatives., Comment: main body: 9 pages, supplemental material: 7 pages. This version corrected a few typos in previous version

The problem studied in this paper is ultrasound image reconstruction from frequency-domain measurements of the scattered field from an object with contrast in attenuation and sound speed. The case where the object has uniform but unknown contrast in these properties relative to the background is considered. Background clutter is taken into account in a physically realistic manner by considering an exact scattering model for randomly located small scatterers that vary in sound speed. The resulting statistical characteristics of the interference is incorporated into the imaging solution, which includes applying a total-variation minimization based approach where the relative effect of perturbation in sound speed to attenuation is included as a parameter. Convex optimization methods provide the basis for the reconstruction algorithm. Numerical data for inversion examples are generated by solving the discretized Lippman-Schwinger equation for the object and speckle-forming scatterers in the background. A statistical model based on the Born approximation is used for reconstruction of the object profile. Results are presented for a two dimensional problem in terms of classification performance and compared to minimum-l2-norm reconstruction. Classification using the proposed method is shown to be robust down to a signal-to-clutter ratio of less than 1 dB., Comment: 25 pages, 10 figures

In this research, we are considering the use of the inverse perspective transformation in video surveillance applications that observe (and possible influence) scenes consisting of moving and stationary objects; e.g., people on a parking area. In previous research, objects were detected on video streams and identified as moving or stationary. Subsequently, distance maps were generated by the Fast Exact Euclidean Distance (FEED) transformation, which uses frame-to-frame information to generate distance maps for video frames in a fast manner. From the resulting distance maps, different kinds of surveillance parameters can be derived. The camera was placed above the scene, and hence, no inverse perspective transformation was needed. In this work,the case is considered the case that the camera is placed under an arbitrary angle on the side of the scene, which might be a more feasible placement than on the top. It will be shown that an image taken from a camera on the side can be easily and fast converted to an image as would be taken by a camera on the top. The allows the use of the previously developed methods after converting each frame of a video stream or only objects of interest detected on them.

Photometry of astrophysical sources, galaxies and stars, in crowded field images, if an old problem, is still a challenging goal, as new space survey missions are launched, releasing new data with increased sensibility, resolution and field of view. The GALEX mission, observes in two UV bands and produces deep sky images of millions of galaxies or stars mixed together. These UV observations are of lower resolution than same field observed in visible bands, and with a very faint signal, at the level of the photon noise for a substantial fraction of objects. Our purpose is to use the better known optical counterparts as prior information in a Bayesian approach to deduce the UV flux. Photometry of extended sources has been addressed several times using various techniques: background determination via sigma clipping, adaptative-aperture, point-spread-function photometry, isophotal photometry, to lists some. The Bayesian approach of using optical priors for solving the UV photometry has already been applied by our team in a previous work. Here we describe the improvement of using the extended shape inferred by deblending the high resolution optical images and not only the position of the optical sources. The resulting photometric accuracy has been tested with simulation of crowded UV fields added on top of real UV images. Finally, this helps to converge to smaller and flat residual and increase the faint source detection threshold. It thus gives the opportunity to work on 2nd order effects, like improving the knowledge of the background or point-spread function by iterating on them.

We present a new method for multi-object segmentation in a maximum a posteriori estimation framework. Our method is motivated by the observation that neighboring or coupling objects in images generate configurations and co-dependencies which could potentially aid in segmentation if properly exploited. Our approach employs coupled shape and inter-shape pose priors that are computed using training images in a nonparametric multi-variate kernel density estimation framework. The coupled shape prior is obtained by estimating the joint shape distribution of multiple objects and the inter-shape pose priors are modeled via standard moments. Based on such statistical models, we formulate an optimization problem for segmentation, which we solve by an algorithm based on active contours. Our technique provides significant improvements in the segmentation of weakly contrasted objects in a number of applications. In particular for medical image analysis, we use our method to extract brain Basal Ganglia structures, which are members of a complex multi-object system posing a challenging segmentation problem. We also apply our technique to the problem of handwritten character segmentation. Finally, we use our method to segment cars in urban scenes.

The theory of compressive sensing (CS) enables the reconstruction of a sparse or compressible image or signal from a small set of linear, non-adaptive (even random) projections. However, in many applications, including object and target recognition, we are ultimately interested in making a decision about an image rather than computing a reconstruction. We propose here a framework for compressive classification that operates directly on the compressive measurements without first reconstructing the image. We dub the resulting dimensionally reduced matched filter the smashed filter. The first part of the theory maps traditional maximum likelihood hypothesis testing into the compressive domain; we find that the number of measurements required for a given classification performance level does not depend on the sparsity or compressibility of the images but only on the noise level. The second part of the theory applies the generalized maximum likelihood method to deal with unknown transformations such as the translation, scale, or viewing angle of a target object. We exploit the fact the set of transformed images forms a low-dimensional, nonlinear manifold in the high-dimensional image space. We find that the number of measurements required for a given classification performance level grows linearly in the dimensionality of the manifold but only logarithmically in the number of pixels/samples and image classes. Using both simulations and measurements from a new single-pixel compressive camera, we demonstrate the effectiveness of the smashed filter for target classification using very few measurements.

Can we recover a signal f ∈R N from a small number of linear measurements? A series of recent papers developed a collection of results showing that it is surprisingly possible to reconstruct certain types of signals accurately from limited measurements. In a nutshell, suppose that f is compressible in the sense that it is well-approximated by a linear combination of M vectors taken from a known basis Ψ. Then not knowing anything in advance about the signal, f can (very nearly) be recovered from about M log N generic nonadaptive measurements only. The recovery procedure is concrete and consists in solving a simple convex optimization program. In this paper, we show that these ideas are of practical significance. Inspired by theoretical developments, we propose a series of practical recovery procedures and test them on a series of signals and images which are known to be well approximated in wavelet bases. We demonstrate that it is empirically possible to recover an object from about 3 M -5 M projections onto generically chosen vectors with an accuracy which is as good as that obtained by the ideal M -term wavelet approximation. We briefly discuss possible implications in the areas of data compression and medical imaging.

The aim of seismic imaging is to reconstruct subsurface reflectivity from scattered acoustic data. In standard reconstruction techniques, the reflectivity model parameters are usually defined as a grid of point scatterers over the area or volume of the subsurface to be imaged. We propose an approach to subsurface imaging using the Dual Tree Complex Wavelet Transform (DT-CWT) as a basis for the reflectivity. This basis is used in conjunction with an iterative optimization which frames the problem as a linearized inverse scattering problem. We demonstrate the method on synthetic data and a marine seismic data set acquired over the Gippsland Basin near Australia. The technique is shown to reduce noise and processing artifacts while preserving discontinuities. It is likely to be particularly useful in cases where the acquired date is incomplete.