Mixture Model-based Approach Research Articles

Automatic detection of nasalization in speech via cue-based analysis

Modern speech recognition systems are powerful, but it is not simple to understand the mechanisms by which they model speech, or precisely how they make use of acoustic cues. Moreover, a system for the explicit detection of individual acoustic cues may be useful not only for speech recognition but for understanding the principles that govern sub-phonemic patterns of surface-phonetic variation, and how specific cue patterns may reflect speech disorders. This report describes a transparent, acoustic cue-based nasalization detection module, that can be used to find not only nasal consonants but also nasalization wherever it appears, e.g., within a vocalic region. We lay out a framework for extracting measurements that are key to detecting nasalization, in particular, the nasality-related spectral peaks P0 and P1, and confirm the utility of these measurements for nasality detection in spoken words. In particular, we present a pre-processing and cue-value extraction framework, and propose a Gaussian Mixture Model-based approach to detect the regions of nasalization in a speech signal. This work on a detection module for nasalization is part of a larger effort to develop a speech recognition system that is based on landmarks and other acoustic cues.

Identification of New Clusters from Labeled Data Using Mixture Models.

Nowadays attempts to segment classes or groups are often found in various fields. Especially, one of emerging issues in biological and medical areas is identification of new subtypes of biological samples or patients. For the identification, we often need to find new subtypes from known classes. In such cases, we usually use clustering techniques. However, usual clustering methods could mix up the labels of the known classes in clustering outcomes and it might lead to wrong interpretation for the identified clusters. Also, they do not use the information about known classes. Thus, this study proposes a Gaussian mixture model-based approach for identifying new clusters from known classes while it maintains them. The performance of the proposed model is verified through simulations and it is applied to a breast cancer data set.

Entropy-Based Variational Scheme with Component Splitting for the Efficient Learning of Gamma Mixtures.

Finite Gamma mixture models have proved to be flexible and can take prior information into account to improve generalization capability, which make them interesting for several machine learning and data mining applications. In this study, an efficient Gamma mixture model-based approach for proportional vector clustering is proposed. In particular, a sophisticated entropy-based variational algorithm is developed to learn the model and optimize its complexity simultaneously. Moreover, a component-splitting principle is investigated, here, to handle the problem of model selection and to prevent over-fitting, which is an added advantage, as it is done within the variational framework. The performance and merits of the proposed framework are evaluated on multiple, real-challenging applications including dynamic textures clustering, objects categorization and human gesture recognition.

Entropy-based Variational Learning of Finite Inverted Beta-Liouville Mixture Model

Mixture models are a common unsupervised learning technique that have been widely used to statistically approximate and analyse heterogenous data. In this paper, an effective mixture model-based approach for positive vectors clustering and modeling is proposed. Our mixture model is based on the inverted Beta-Liouville (IBL) distribution. To deploy the proposed model, we introduce an entropy-based variational inference algorithm. The performance of the proposed model is evaluated on two real-world applications, namely, human activity recognition and image categorization.

IonQuant Enables Accurate and Sensitive Label-Free Quantification With FDR-Controlled Match-Between-Runs

Missing values weaken the power of label-free quantitative proteomic experiments to uncover true quantitative differences between biological samples or experimental conditions. Match-between-runs (MBR) has become a common approach to mitigate the missing value problem, where peptides identified by tandem mass spectra in one run are transferred to another by inference based on m/z, charge state, retention time, and ion mobility when applicable. Though tolerances are used to ensure such transferred identifications are reasonably located and meet certain quality thresholds, little work has been done to evaluate the statistical confidence of MBR. Here, we present a mixture model-based approach to estimate the false discovery rate (FDR) of peptide and protein identification transfer, which we implement in the label-free quantification tool IonQuant. Using several benchmarking datasets generated on both Orbitrap and timsTOF mass spectrometers, we demonstrate superior performance of IonQuant with FDR-controlled MBR compared with MaxQuant (19–38 times faster; 6–18% more proteins quantified and with comparable or better accuracy). We further illustrate the performance of IonQuant and highlight the need for FDR-controlled MBR, in two single-cell proteomics experiments, including one acquired with the help of high-field asymmetric ion mobility spectrometry separation. Fully integrated in the FragPipe computational environment, IonQuant with FDR-controlled MBR enables fast and accurate peptide and protein quantification in label-free proteomics experiments.

SAME-clustering: Single-cell Aggregated Clustering via Mixture Model Ensemble.

Clustering is an essential step in the analysis of single cell RNA-seq (scRNA-seq) data to shed light on tissue complexity including the number of cell types and transcriptomic signatures of each cell type. Due to its importance, novel methods have been developed recently for this purpose. However, different approaches generate varying estimates regarding the number of clusters and the single-cell level cluster assignments. This type of unsupervised clustering is challenging and it is often times hard to gauge which method to use because none of the existing methods outperform others across all scenarios. We present SAME-clustering, a mixture model-based approach that takes clustering solutions from multiple methods and selects a maximally diverse subset to produce an improved ensemble solution. We tested SAME-clustering across 15 scRNA-seq datasets generated by different platforms, with number of clusters varying from 3 to 15, and number of single cells from 49 to 32 695. Results show that our SAME-clustering ensemble method yields enhanced clustering, in terms of both cluster assignments and number of clusters. The mixture model ensemble clustering is not limited to clustering scRNA-seq data and may be useful to a wide range of clustering applications.

Modeling and Clustering Positive Vectors via Nonparametric Mixture Models of Liouville Distributions.

In this article, we propose an effective mixture model-based approach to modeling and clustering positive data vectors. Our mixture model is based on the inverted Beta-Liouville (IBL) distribution which is extracted from the family of Liouville distributions. To cope with the problem of determining the appropriate number of clusters in our approach, a nonparametric Bayesian framework is used to extend the IBL mixture to an infinite mixture model in which the number of clusters is assumed to be infinite initially and will be inferred automatically during the learning process. To optimize the proposed model, we propose a convergence-guaranteed learning algorithm based on the averaged collapsed variational Bayes inference that can effectively learn model parameters with closed-form solutions. The effectiveness of the proposed infinite IBL mixture model for modeling and clustering positive vectors is validated through both synthetic and real-world data sets.

Flexible High-Dimensional Unsupervised Learning with Missing Data.

The mixture of factor analyzers (MFA) model is a famous mixture model-based approach for unsupervised learning with high-dimensional data. It can be useful, inter alia, in situations where the data dimensionality far exceeds the number of observations. In recent years, the MFA model has been extended to non-Gaussian mixtures to account for clusters with heavier tail weight and/or asymmetry. The generalized hyperbolic factor analyzers (MGHFA) model is one such extension, which leads to a flexible modelling paradigm that accounts for both heavier tail weight and cluster asymmetry. In many practical applications, the occurrence of missing values often complicates data analyses. A generalization of the MGHFA is presented to accommodate missing values. Under a missing-at-random mechanism, we develop a computationally efficient alternating expectation conditional maximization algorithm for parameter estimation of the MGHFA model with different patterns of missing values. The imputation of missing values under an incomplete-data structure of MGHFA is also investigated. The performance of our proposed methodology is illustrated through the analysis of simulated and real data.

A Dirichlet-multinomial mixture model-based approach for daily solar radiation classification

A challenging problem in the classification of daily solar radiation is the selection of the appropriate model complexity and size that best describe the data. This paper introduces a new nonparametric Bayesian method for automatic classification of daily clearness index, by assuming Dirichlet process as a nonparametric prior on the model parameters. Nonparametric methods are free from the parametric model assumptions, and there is no need to specify any parametric specifications, or to restrict the number of classes. Our approach relies on the inference of the posterior distributions using the collapsed Gibbs sampler. The proposed method is tested using measurements from 2003 to 2016, at the Silver Lake monitoring station in the USA (121°3′W, 43°7′N), with a 5-min logging interval. By applying our classification algorithm, three classes of daily clearness index distributions are identified, corresponding to three types of sky cloudiness, namely cloudy, partially cloudy, and clear sky. The proposed classification framework can facilitate the design of solar radiation conversion systems.

Smart healthcare monitoring: a voice pathology detection paradigm for smart cities

With the increasing demand for automated, remote, intelligent, and real-time healthcare services in smart cities, smart healthcare monitoring is necessary to provide improved and complete care to residents. In this monitoring, health-related media or signals collected from smart-devices/objects are transmitted and processed to cater to the need for quality care. However, it is challenging to create a framework or method to handle media-related healthcare data analytics or signals (e.g., voice/audio, video, or electroglottographic (EGG) signals) to meet the complex on-demand healthcare needs for successful smart city management. To this end, this paper proposes a cloud-oriented smart healthcare monitoring framework that interacts with surrounding smart devices, environments, and smart city stakeholders for affordable and accessible healthcare. As a smart city healthcare monitoring case study, a voice pathology detection (VPD) method is proposed. In the proposed method, two types of input, a voice signal and an EGG signal, are used. The input devices are connected to the Internet and the captured signals are transmitted to the cloud. The signals are then processed and classified as either normal or pathologic with a confidence score. These results are passed to registered doctors that make the final decision and take appropriate action. To process the signals, local features are extracted from the first-order derivative of the voice signal, and shape and cepstral features are extracted from the EGG signal. For classification, a Gaussian mixture model-based approach is used. Experimental results show that the proposed method can achieve VPD that is more than 93% accurate.

Mixture model-based association analysis with case-control data in genome wide association studies.

Multilocus haplotype analysis of candidate variants with genome wide association studies (GWAS) data may provide evidence of association with disease, even when the individual loci themselves do not. Unfortunately, when a large number of candidate variants are investigated, identifying risk haplotypes can be very difficult. To meet the challenge, a number of approaches have been put forward in recent years. However, most of them are not directly linked to the disease-penetrances of haplotypes and thus may not be efficient. To fill this gap, we propose a mixture model-based approach for detecting risk haplotypes. Under the mixture model, haplotypes are clustered directly according to their estimated disease penetrances. A theoretical justification of the above model is provided. Furthermore, we introduce a hypothesis test for haplotype inheritance patterns which underpin this model. The performance of the proposed approach is evaluated by simulations and real data analysis. The results show that the proposed approach outperforms an existing multiple testing method.

A mixture model-based nonparametric approach to estimating a count distribution

The density of a mixture distribution with unknown discrete mixing distribution can be a way of finding a nonparametric estimate of a density. Comparing with a standard parametric approach that runs the risk of model misspecification and a kernel-based nonparametric approach that retains all data points for constructing a density estimate, this mixture model-based nonparametric approach that naturally circumvents these problems is very appealing to density estimation. Owing to these advantages, this approach to estimating a count distribution is considered. Estimation of a count distribution via this means necessarily involves the problem of nonparametric estimation of a discrete mixing distribution. The nonparametric estimation problem is formulated using the family of power divergences, thus offering a class of nonparametric estimators of a discrete mixing distribution. A fast algorithm is presented for computing the class of nonparametric minimum power divergence estimates with which the corresponding class of mixture model-based estimates of a count distribution can be implemented. Only the use of the probability mass function of the Poisson mixture distribution with unknown discrete mixing distribution is illustrated. The simulation results show that the mixture model-based estimators on average outperform the kernel-based estimators for uncontaminated samples of larger sizes and contaminated samples of all sizes considered.

Mitigation of Pre-existing Antibodies to a Biotherapeutic in Non-clinical Species When Establishing Anti-drug Antibody Assay Cutpoint.

Biotherapeutics are known for their potential to induce drug specific immune responses, which are commonly evaluated by the detection of anti-drug antibodies (ADAs). For some biotherapeutics, pre-existing ADAs against drug have been observed in drug-naïve matrix. The presence of pre-existing drug specific antibodies may significantly complicate assessment of the screening ADA assay cutpoint value, which is usually established based on the statistical analysis of signal distribution from the drug-naïve individuals. A Gaussian mixture model-based approach is presented herein to address high prevalence of pre-existing ADAs to a modified monoclonal antibody-based biotherapeutic (m-mAb). A high prevalence of pre-existing anti-m-mAb antibodies was observed in drug-naïve individual cynomolgus monkey serum samples with signal ranging from 100 to 7000 relative light units (RLU, as determined in an electrochemiluminescence readout-based assay). Application of the industry standard statistical algorithm resulted in a relatively high floating screening assay cutpoint factor (CPF) of 9.80, which potentially would have reported a high percent of false negative samples. An alternative, Gaussian mixture model-based approach was applied to identify the least reactive individual samples in the tested population, which resulted in a floating screening assay CPF of 2.35. The low CPF value significantly reduced the risk of reporting false negative results. The proposed Gaussian mixture model-based approach described herein provides an alternate method for the calculation of biologically relevant screening assay CPF when high prevalence of pre-existing drug specific antibodies is observed.

Piecewise Regression Mixture for Simultaneous Functional Data Clustering and Optimal Segmentation

This paper introduces a novel mixture model-based approach to the simultaneous clustering and optimal segmentation of functional data, which are curves presenting regime changes. The proposed model consists of a finite mixture of piecewise polynomial regression models. Each piecewise polynomial regression model is associated with a cluster, and within each cluster, each piecewise polynomial component is associated with a regime (i.e., a segment). We derive two approaches to learning the model parameters: the first is an estimation approach which maximizes the observed-data likelihood via a dedicated expectation-maximization (EM) algorithm, then yielding a fuzzy partition of the curves into K clusters obtained at convergence by maximizing the posterior cluster probabilities. The second is a classification approach and optimizes a specific classification likelihood criterion through a dedicated classification expectation-maximization (CEM) algorithm. The optimal curve segmentation is performed by using dynamic programming. In the classification approach, both the curve clustering and the optimal segmentation are performed simultaneously as the CEM learning proceeds. We show that the classification approach is a probabilistic version generalizing the deterministic K-means-like algorithm proposed in Hebrail, Hugueney, Lechevallier, and Rossi (2010). The proposed approach is evaluated using simulated curves and real-world curves. Comparisons with alternatives including regression mixture models and the K-means-like algorithm for piecewise regression demonstrate the effectiveness of the proposed approach.

Clustering, classification, discriminant analysis, and dimension reduction via generalized hyperbolic mixtures

A method for dimension reduction with clustering, classification, or discriminant analysis is introduced. This mixture model-based approach is based on fitting generalized hyperbolic mixtures on a reduced subspace within the paradigm of model-based clustering, classification, or discriminant analysis. A reduced subspace of the data is derived by considering the extent to which group means and group covariances vary. The members of the subspace arise through linear combinations of the original data, and are ordered by importance via the associated eigenvalues. The observations can be projected onto the subspace, resulting in a set of variables that captures most of the clustering information available. The use of generalized hyperbolic mixtures gives a robust framework capable of dealing with skewed clusters. Although dimension reduction is increasingly in demand across various application areas, many applications are biological and so some of the real data examples are within that sphere. Simulated data are also used for illustration.

Identifying Authorities in Online Communities

Several approaches have been proposed for the problem of identifying authoritative actors in online communities. However, the majority of existing methods suffer from one or more of the following limitations: (1) There is a lack of an automatic mechanism to formally discriminate between authoritative and nonauthoritative users. In fact, a common approach to authoritative user identification is to provide a ranked list of users expecting authorities to come first. A major problem of such an approach is the question of where to stop reading the ranked list of users. How many users should be chosen as authoritative? (2) Supervised learning approaches for authoritative user identification suffer from their dependency on the training data. The problem here is that labeled samples are more difficult, expensive, and time consuming to obtain than unlabeled ones. (3) Several approaches rely on some user parameters to estimate an authority score. Detection accuracy of authoritative users can be seriously affected if incorrect values are used. In this article, we propose a parameterless mixture model-based approach that is capable of addressing the three aforementioned issues in a single framework. In our approach, we first represent each user with a feature vector composed of information related to its social behavior and activity in an online community. Next, we propose a statistical framework, based on the multivariate beta mixtures, in order to model the estimated set of feature vectors. The probability density function is therefore estimated and the beta component that corresponds to the most authoritative users is identified. The suitability of the proposed approach is illustrated on real data extracted from the Stack Exchange question-answering network and Twitter.

Human object annotation for surveillance video forensics

A system that can automatically annotate surveillance video in a manner useful for locating a person with a given description of clothing is presented. Each human is annotated based on two appearance features: primary colors of clothes and the presence of text/logos on clothes. The annotation occurs after a robust foreground extraction stage employing a modified Gaussian mixture model-based approach. The proposed pipeline consists of a preprocessing stage where color appearance of an image is improved using a color constancy algorithm. In order to annotate color information for human clothes, we use the color histogram feature in HSV space and find local maxima to extract dominant colors for different parts of a segmented human object. To detect text/logos on clothes, we begin with the extraction of connected components of enhanced horizontal, vertical, and diagonal edges in the frames. These candidate regions are classified as text or nontext on the basis of their local energy-based shape histogram features. Further, to detect humans, a novel technique has been proposed that uses contourlet transform-based local binary pattern (CLBP) features. In the proposed method, we extract the uniform direction invariant LBP feature descriptor for contourlet transformed high-pass subimages from vertical and diagonal directional bands. In the final stage, extracted CLBP descriptors are classified by a trained support vector machine. Experimental results illustrate the superiority of our method on large-scale surveillance video data.

Tracking continuous emotional trends of participants during affective dyadic interactions using body language and speech information

We address the problem of tracking continuous levels of a participant's activation, valence and dominance during the course of affective dyadic interactions, where participants may be speaking, listening or doing neither. To this end, we extract detailed and intuitive descriptions of each participant's body movements, posture and behavior towards his interlocutor, and speech information. We apply a Gaussian Mixture Model-based approach which computes a mapping from a set of observed audio–visual cues to an underlying emotional state. We obtain promising results for tracking trends of participants' activation and dominance values, which outperform other regression-based approaches used in the literature. Additionally, we shed light into the way expressive body language is modulated by underlying emotional states in the context of dyadic interactions.

Detection of allele-specific methylation through a generalized heterogeneous epigenome model

Motivations: High-throughput sequencing has made it possible to sequence DNA methylation of a whole genome at the single-base resolution. A sample, however, may contain a number of distinct methylation patterns. For instance, cells of different types and in different developmental stages may have different methylation patterns. Alleles may be differentially methylated, which may partially explain that the large portions of epigenomes from single cell types are partially methylated, and may have major effects on transcriptional output. Approaches relying on DNA sequence polymorphism to identify individual patterns from a mixture of heterogeneous epigenomes are insufficient as methylcytosines occur at a much higher density than SNPs.Results: We have developed a mixture model-based approach for resolving distinct epigenomes from a heterogeneous sample. In particular, the model is applied to the detection of allele-specific methylation (ASM). The methods are tested on a synthetic methylome and applied to an Arabidopsis single root cell methylome.Contact: qpeng@cs.ucsd.edu

A phylogenetic mixture model for the identification of functionally divergent protein residues

To understand the evolution of molecular function within protein families, it is important to identify those amino acid residues responsible for functional divergence; i.e. those sites in a protein family that affect cofactor, protein or substrate binding preferences; affinity; catalysis; flexibility; or folding. Type I functional divergence (FD) results from changes in conservation (evolutionary rate) at a site between protein subfamilies, whereas type II FD occurs when there has been a shift in preferences for different amino acid chemical properties. A variety of methods have been developed for identifying both site types in protein subfamilies, both from phylogenetic and information-theoretic angles. However, evaluation of the performance of these methods has typically relied upon a handful of reasonably well-characterized biological datasets or analyses of a single biological example. While experimental validation of many truly functionally divergent sites (true positives) can be relatively straightforward, determining that particular sites do not contribute to functional divergence (i.e. false positives and true negatives) is much more difficult, resulting in noisy 'gold standard' examples. We describe a novel, phylogeny-based functional divergence classifier, FunDi. Unlike previous approaches, FunDi uses a unified mixture model-based approach to detect type I and type II FD. To assess FunDi's overall classification performance relative to other methods, we introduce two methods for simulating functionally divergent datasets. We find that the FunDi method performs better than several other predictors over a wide variety of simulation conditions. http://rogerlab.biochem.dal.ca/Software andrew.roger@dal.ca Supplementary data are available at Bioinformatics online.