Mixture models

In this paper we present a framework for realtime online signature verification scenarios. The proposed framework is based on state-of-the-art feature extraction and Gaussian Mixture Model (GMM) classification. While our signature verification library is generally applicable to any input device using digital pens, we have implemented verification scenarios using the Anoto digital pen. As such our automated signature verification framework becomes an interesting commodity for industry, because the Anoto SDK is easy to apply and the GMM-based classification can be seamlessly integrated. The novelty of this work is the application of our framework that takes real-time online signature verification to every scenario where digital pens may potentially be used. In this paper we describe several scenarios where our framework has been applied, including signatures in financial contracts or ordering processes. We also propose a general approach to integrate the GMM-descriptions into electronic ID-cards in order to also store behavioral biometrics on these cards. In experiments we have measured the performance of the signature verification system when skilled forgeries were present. The interest shown by our partner financial institutions and the results of our initial evaluations indicate that our signature verification framework suits exactly the demands of our clients.

Early heart disease control can be achieved by high disease prediction and diagnosis efficiency. This paper focuses on the use of model based clustering techniques to predict and diagnose heart disease via Poisson mixture regression models. Analysis and application of Poisson mixture regression models is here addressed under two different classes: standard and concomitant variable mixture regression models. Results show that a two-component concomitant variable Poisson mixture regression model predicts heart disease better than both the standard Poisson mixture regression model and the ordinary general linear Poisson regression model due to its low Bayesian Information Criteria value. Furthermore, a Zero Inflated Poisson Mixture Regression model turned out to be the best model for heart prediction over all models as it both clusters individuals into high or low risk category and predicts rate to heart disease componentwise given clusters available. It is deduced that heart disease prediction can be effectively done by identifying the major risks componentwise using Poisson mixture regression model.

In many long-term clinical trials subjects often experience a number of events all of which might serve as important endpoints for medical studies. The analysis of such multiple events can be beneficial from both medical and statistical perspectives. We review and discuss some recent developments in the area of Bayesian methodologies for the analysis of multiple event times data, focusing more on semiparametric models based on the intensity functions. The nonparametric part of such models is modeled as a realization of a stochastic process; while the parametric part, which may include a regression parameter and a parameter quantifying the heterogeneity (i.e., frailty) of a population, is assumed to have a parametric distribution with possibly unknown hyperparameters. Numerical computations required to obtain posterior estimates of the parameters are generally based on a class of sampling based methods, broadly known as the Markov Chain Monte Carlo (MCMC) algorithms. The methodologies discussed in this chapter are motivated by and aimed at analyzing some common types of multiple event times data that are often observed in medical studies. Bayesian Exploratory Data Analysis (EDA) methods and diagnostics for model selection and model assessment are considered for each of these cases.

N-gram language modeling typically requires large quantities of in-domain training data, i.e., data that matches the task in both topic and style. For conversational speech applications, particularly meeting transcription, obtaining large volumes of speech transcripts is often unrealistic; topics change frequently and collecting conversational-style training data is time-consuming and expensive. In particular, new topics introduce new vocabulary items which are not included in existing models. In this work, we use a variety of data sources (reflecting different sizes and styles), combined using mixture n-gram models. We study the impact of the different sources on vocabulary expansion and recognition accuracy, and investigate possible indicators of the usefulness of a data source. For the task of recognizing meeting speech, we obtain a 9% relative reduction in the overall word error rate and a 61% relative reduction in the word error rate for "new" words added to the vocabulary over a baseline language model trained from general conversational speech data.

Large-scale multilocus studies have become common in molecular phylogenetics, but the best way to interpret these studies when their results strongly conflict with prior information about phylogeny remains unclear. An example of such a conflict is provided by the ratites (the large flightless birds of southern land masses, including ostriches, emus, and rheas). Ratite monophyly is strongly supported by both morphological data and many earlier molecular studies and is used as a textbook example of vicariance biogeography. However, recent studies have indicated that ratites are not monophyletic; instead, the volant tinamous nest inside the ratites rather than forming their sister group within the avian superorder Palaeognathae. Large-scale studies can exhibit biases that reflect a number of factors, including limitations in the fit of the evolutionary models used for analyses and problems with sequence alignment, so the unexpected conclusion that ratites are not monophyletic needs to ...

In this paper we study how perturbing a matrix changes its non-negative rank. We prove that the non-negative rank is upper-semicontinuos and we describe some special families of perturbations. We show how our results relate to Statistics in terms of the study of Maximum Likelihood Estimation for mixture models.

x Our comments about the paper by Leeflang and Wittink Internat. J. Res. Marketing, 17 2000 105 comprise of two components: first, we address two issues on which we disagree with Leeflang and Wittink: soft versus hard data, and individual-level versus segment-level models. Secondly, we supplement their paper by attempting to predict how marketing model building will change in the future. q

We consider the problem of feature-based face recognition in the setting where only a single example of each face is available for training. The mixture-distance technique we introduce achieves a recognition rate of 95% on a database of 685 people in which each face is represented by 30 measured distances. This is currently the best recorded recognition rate for a feature-based system applied to a database of this size. By comparison, nearest neighbor search using Euclidean distance yields 84%. In our work a novel distance function is constructed based on local second order statistics as estimated by modeling the training data as a mixture of normal densities. We report on the results from mixtures of several sizes. We demonstrate that a at mixture of mixtures performs as well as the best model and therefore represents an e ective solution to the model selection problem. A mixture perspective is also taken for individual Gaussians to choose between rst order (variance) and second order (covariance) models. Here an approximation to at combination is proposed and seen to perform well in practice. Our results demonstrate that even in the absence of multiple training examples for each class, it is sometimes possible to infer from a statistical model of training data, a signi cantly improved distance function for use in pattern recognition.

An unsupervised learning algorithm that can obtain a probabilistic model of an object composed of a collection of parts (a moving human body in our examples) automatically from unlabeled training data is presented. The training data include both useful "foreground" features as well as features that arise from irrelevant background clutter-the correspondence between parts and detected features is unknown. The joint probability density function of the parts is represented by a mixture of decomposable triangulated graphs which allow for fast detection. To learn the model structure as well as model parameters, an EM-like algorithm is developed where the labeling of the data (part assignments) is treated as hidden variables. The unsupervised learning technique is not limited to decomposable triangulated graphs. The efficiency and effectiveness of our algorithm is demonstrated by applying it to generate models of human motion automatically from unlabeled image sequences, and testing the learned models on a variety of sequences.

In many contexts, warning systems of law enforcement are used to let uninformed individuals learn what is illegal, while sanctions are applied only after a number of repeated violations. Surprisingly no em-pirical evidence is available so far, over the learning impact of warnings. This paper is a first attempt to empirically investigate the warning's effect on individuals' behavior employing a unique database on a traffic law enforcement system, which constitutes an extraordinary nat-ural laboratory to test whether experience warning induces learning. Specifically, we use six-year longitudinal data on about 50000 drivers under the Italian point-record system of traffic law. Our statistical re-sults show that warned drivers become more compliant. To the extent individuals learn through their repeated behavior, a warning system makes it possible to apply sanctions only to (presumably) informed violators.

The WHO Collaborating Centre for International Drug Monitoring in Uppsala, Sweden, maintains and analyses the world's largest database of reports on suspected adverse drug reaction incidents that occur after drugs are introduced on the market. As in other post-marketing drug safety data sets, the presence of duplicate records is an important data quality problem and the detection of duplicates in the WHO drug safety database remains a formidable challenge, especially since the reports are anonymised before submitted to the database. However, to our knowledge no work has been published on methods for duplicate detection in postmarketing drug safety data. In this paper, we propose a method for probabilistic duplicate detection based on the hit-miss model for statistical record linkage described by Copas & Hilton. We present two new generalisations of the standard hit-miss model: a hit-miss mixture model for errors in numerical record fields and a new method to handle correlated record fields. We demonstrate the effectiveness of the hit-miss model for duplicate detection in the WHO drug safety database both at identifying the most likely duplicate for a given record (94.7% accuracy) and at discriminating duplicates from random matches (63% recall with 71% precision). The proposed method allows for more efficient data cleaning in post-marketing drug safety data sets, and perhaps other applications throughout the KDD community.

T he mixed or heterogeneous multinomial logit (MIXL) model has become popular in a number of fields, especially marketing, health economics, and industrial organization. In most applications of the model, the vector of consumer utility weights on product attributes is assumed to have a multivariate normal (MVN) distribution in the population. Thus, some consumers care more about some attributes than others, and the IIA property of multinomial logit (MNL) is avoided (i.e., segments of consumers will tend to switch among the subset of brands that possess their most valued attributes). The MIXL model is also appealing because it is relatively easy to estimate. Recently, however, some researchers have argued that the MVN is a poor choice for modelling taste heterogeneity. They argue that much of the heterogeneity in attribute weights is accounted for by a pure scale effect (i.e., across consumers, all attribute weights are scaled up or down in tandem). This implies that choice behaviour is simply more random for some consumers than others (i.e., holding attribute coefficients fixed, the scale of their error term is greater). This leads to a "scale heterogeneity" MNL model (S-MNL). Here, we develop a generalized multinomial logit model (G-MNL) that nests S-MNL and MIXL. By estimating the S-MNL, MIXL, and G-MNL models on 10 data sets, we provide evidence on their relative performance. We find that models that account for scale heterogeneity (i.e., G-MNL or S-MNL) are preferred to MIXL by the Bayes and consistent Akaike information criteria in all 10 data sets. Accounting for scale heterogeneity enables one to account for "extreme" consumers who exhibit nearly lexicographic preferences, as well as consumers who exhibit very "random" behaviour (in a sense we formalize below).

An important goal of microarray studies is the detection of genes that show significant changes in observed expressions when two or more classes of biological samples such as treatment and control are compared. Using the c-fold rule, a gene is declared to be differentially expressed if its average expression level varies by more than a constant factor c between treatment and control (typically c = 2). While often used, however, this simple rule is not completely convincing. We propose to model this filter and define a binary variable at the gene×experiment level, allowing for a more powerful treatment of the corresponding information. We introduce a gene-specific random term controlling for both dependence among genes and variability with respect to the c-fold threshold. We make inference via a two-level finite mixture model under a likelihood approach. Then, using the counting distribution we show how parameter estimates can be usefully derived also under a Bayesian nonparametric approach which allows to keep under control some error rate of erroneous discoveries. We illustrate the effectiveness of both proposed approaches through a largescale simulation study and an real-data application based on the well known dataset introduced by Alon et al. (1999).

This paper estimates the risk preferences of cotton farmers in Southern Peru, using the results from a multiple-price-list lottery game. Assuming that preferences conform to two of the leading models of decision under risk|Expected Utility Theory (EUT) and Cumulative Prospect Theory (CPT)|we nd strong evidence of moderate risk aversion. Once we include individual characteristics in the estimation of risk parameters, we observe that farmers use subjective nonlinear probability weighting, a behavior consistent with CPT. Interestingly, when we allow for preference heterogeneity via the estimation of mixture models|where the proportion of subjects who behave according to EUT or to CPT is endogenously determined|we nd that the majority of farmers' choices are best explained by CPT. We further hypothesize that the multiple switching behavior observed in our sample can be explained by nonlinear probability weighting made in a context of large random calculation mistakes; the evidence found on this regard is mixed. Finally, we nd that attaining higher education is the single most important individual characteristic correlated with risk preferences, a result that suggests a connection between cognitive abilities and behavior towards risk.

The objective is to calculate the probability, P F , that a device will fail when its inputs, x, are randomly distributed with probability density, p (x), e.g., the probability that a device will fracture when subject to varying loads. Here failure is defined as some scalar function, y (x), exceeding a threshold, T. If evaluating y (x) via physical or numerical experiments is sufficiently expensive or P F is sufficiently small, then Monte Carlo (MC) methods to estimate P F will be unfeasible due to the large number of function evaluations required for a specified accuracy. Importance sampling (IS), i.e., preferentially sampling from "important" regions in the input space and appropriately down-weighting to obtain an unbiased estimate, is one approach to assess P F more efficiently. The inputs are sampled from an importance density, p (x). We present an adaptive importance sampling (AIS) approach which endeavors to adaptively improve the estimate of the ideal importance density, p * (x), during the sampling process. Our approach uses a mixture of component probability densities that each approximate p * (x). An iterative process is used to construct the sequence of improving component probability densities. At each iteration, a Gaussian process (GP) surrogate is used to help identify areas in the space where failure is likely to occur. The GPs are not used to directly calculate the failure probability; they are only used to approximate the importance density. Thus, our Gaussian process adaptive importance sampling (GPAIS) algorithm overcomes limitations involving using a potentially inaccurate surrogate model directly in IS calculations. This robust GPAIS algorithm performs surprisingly well on a pathological test function.

Economists and psychologists have recently been developing new theories of decision making under uncertainty that can accommodate the observed violations of standard statistical decision theoretic axioms by experimental subjects. We propose a procedure that finds a collection of decision rules that best explain the behavior of experimental subjects. The procedure is a combination of maximum likelihood estimation of the rules together with an implicit classification of subjects to the various rules and a penalty for having too many rules. We apply our procedure to data on probabilistic updating by subjects in four different universities. We get remarkably robust results showing that the most important rules used by the subjects (in order of importance) are Bayes's rule, a representativeness rule (ignoring the prior), and, to a lesser extent, conservatism (overweighting the prior).

Future prediction is a fundamental principle of intelligence that helps plan actions and avoid possible dangers. As the future is uncertain to a large extent, modeling the uncertainty and multimodality of the future states is of great relevance. Existing approaches are rather limited in this regard and mostly yield a single hypothesis of the future or, at the best, strongly constrained mixture components that suffer from instabilities in training and mode collapse. In this work, we present an approach that involves the prediction of several samples of the future with a winner-takes-all loss and iterative grouping of samples to multiple modes. Moreover, we discuss how to evaluate predicted multimodal distributions, including the common real scenario, where only a single sample from the ground-truth distribution is available for evaluation. We show on synthetic and real data that the proposed approach triggers good estimates of multimodal distributions and avoids mode collapse.

Wavelet-based demosaicing techniques have the advantage of being computationally relatively fast, while having a reconstruction performance that is similar to state-of-the-art techniques. Because the demosaicing rules are linear, it is fairly simple to integrate denoising into the demosaicing. In this paper, we present a method that performs joint denoising and demosaicing, using a Gaussian Scale Mixture (GSM) prior model, thereby modeling the local edge direction as a hidden variable. The results indicate that this technique offers a better reconstruction performance (in PSNR sense and visually) than sequential demosaicing and denoising. On a recent GPU, our algorithm takes 3.5 s for reconstructing a 12 megapixel RAW digital camera image.

This paper describes a multivariate Poisson mixture model for clustering supermarket shoppers based on their purchase frequency in a set of product categories. The multivariate nature of the model accounts for cross-selling effects between the purchases made in different product categories. However, for computational reasons, most multivariate approaches limit the covariance structure by including just one common interaction term, or by not including any covariance at all. Although this reduces the number of parameters significantly, it is often too simplistic as typically multiple interactions exist on different levels. This paper proposes a theoretically more complete variance/covariance structure of the multivariate Poisson model, based on domain knowledge or preliminary statistical analysis of significant purchase interaction effects in the data. Consequently, the model does not contain more parameters than necessary, whilst still accounting for the existing covariance in the data. Practically, retail category managers can use the model to devise customized merchandising strategies.

Because consumers are limited information processors seeking to conserve cognitive energy, it is likely that at least some use identical decision heuristics across product categories. This study develops a finite mixture logit model that identifies segments of households with identical behaviors across product categories. The proposed model is shown to fit in-sample choices and forecast out-of-sample choices at least as well as an unrestricted model in which all choice behaviors are independent across product categories. The results show that about 32% of the sample households have choice behaviors that are identical across at least two of the three product categories studied, while the remaining households have choice behaviors that are independent across all three categories. The empirical results show that the segment with identical behaviors is quite price sensitive, not at all sensitive to store feature advertising, and not very brand-or size-loyal. These households are more likely to have larger families and marginally lower incomes and to shop less frequently and spend less per shopping trip. They are also lighter users in two of the three product categories investigated. Implications of the model for manufacturers and retailers are discussed. D 2002 Published by Elsevier Science B.V.

The Expectation-Maximization (EM) algorithm is an iterative approach to maximum likelihood parameter estimation. Jordan and Jacobs (1993) recently proposed an EM algorithm for the mixture of experts architecture of Jacobs, Jordan, Nowlan and Hinton (1991) and the hierarchical mixture of experts architecture of Jordan and Jacobs (1992). They showed empirically that the EM algorithm for these architectures yields signi cantly faster convergence than gradient ascent. In the current paper we provide a theoretical analysis of this algorithm. We show that the algorithm can be regarded as a variable metric algorithm with its searching direction having a positive projection on the gradient of the log likelihood. We also analyze the convergence of the algorithm and provide an explicit expression for the convergence rate. In addition, we describe an acceleration technique that yields a signi cant speedup in simulation experiments.

Choice behavior is typically evaluated by assuming that the data is generated by one latent decision-making process or another. What if there are two (or more) latent decision-making processes generating the observed choices? Some choices might then be better characterized as being generated by one process, and other choices by the other process. A finite mixture model can be used to estimate the parameters of each decision process while simultaneously estimating the probability that each process applies to the sample. We consider the canonical case of lottery choices in a laboratory experiment and assume that the data is generated by expected utility theory and prospect theory decision rules. We jointly estimate the parameters of each theory as well as the fraction of choices characterized by each. The methodology provides the wedding invitation, and the data consummates the ceremony followed by a decent funeral for the representative agent model that assumes only one type of decision process. The evidence suggests support for each theory, and goes further to identify under what demographic domains one can expect to see one theory perform better than the other. We therefore propose a reconciliation of the debate over two of the dominant theories of choice under risk, at least for the tasks and samples we consider. The methodology is broadly applicable to a range of debates over competing theories generated by experimental and non-experimental data.

Based on census data linked to household surveys, we analyze the univariate and joint distribution of income, health and education at the municipality level in Mexico from 1990 to 2010 using Gaussian mixture models. The univariate analysis finds an emergence of a low-income cluster in 2000, which disappears again by 2010. Our trivariate estimation shows an education-led human development convergence over time while dynamics are mainly affected by fluctuations in health and income. Changes in development clusters have a clear spatial pattern and are closely related to the relative size of the agricultural sector and the proportion of indigenous population groups.

In this study, the effect of coarse aggregate shape characteristics on the compactability and microstructural properties of asphalt mixtures was virtually investigated using a discreet element method (DEM). Results reveal there is a statistically significant correlation between coarse aggregate flatness, elongation, roundness, and sphericity as shape descriptors and initial asphalt mixture density, as a compaction parameter. Analyses indicate that among all particle shape descriptors, only roundness and regularity had a statistically significant relation with compaction slope, and as the amount of roundness and regularity increase, the compaction slope decreases. Additionally, flatness and elongation values greater than 0.75 result in lower average coordination number values compared to those values less than 0.75. Results also indicate that although the probability density function (PDF) of low contact force highly depends on flatness, elongation and sphericity of particles, the PDF of high and very high contact forces is independent of the particle shape.

Self‐compacting concrete (SCC) was developed in 1988 and introduced by Professor H. Okamura to achieve durability of structures with low skilled laborers. The three properties of SCC that differentiates it from traditional concrete are passing ability, filling ability and segregation resistance. The present study highlights the developments concerning SCC in the period from 1995 to 2017. This information is important for researchers and professionals working in this area. SCC can be optimized to a larger extent to find the best alternative either by use of additives or tailor‐made mix design. Based on different principles and parameters, design methods can be categorized into six classes namely empirical design method, statistical factorial design method, strength‐based design method, rheology of paste method, particle packing method and Eco‐SCC mixture design method. To develop SCC with high quality a suitable method can be chosen depending upon the situation and application.

This paper presents an unsupervised approach for feature selection and extraction in mixtures of generalized Dirichlet (GD) distributions. Our method defines a new mixture model that is able to extract independent and non-Gaussian features without loss of accuracy. The proposed model is learned using the Expectation-Maximization algorithm by minimizing the message length of the data set. Experimental results show the merits of the proposed methodology in the categorization of object images.

This paper is the …rst to apply a …nite mixture model to a sample of 64 nations to endogenously analyze the cross-country growth behavior over the period . Results show that growth patterns were segmented in two worldwide regimes, the one characterized by convergence in per capita income, and the other by divergence. Interestingly, when three historical epochs are distinctly analyzed, in order to investigate the empirical link between globalization and convergence, the dynamics which dominated over the whole period seem to have emerged only during the post-1950 years. In contrast, the First Global Wave was marked by persistent heterogeneities.

Economic historians have stressed that income convergence was a key feature of the 'OECD-club' and that globalization was among the accelerating forces of this process in the long-run. This view has however been challenged, since it suffers from an ad hoc selection of countries. In the paper, a mixture model is applied to a sample of 64 countries to endogenously analyze the cross-country growth behavior over the period 1870-2003. Results show that growth patterns were segmented in two worldwide regimes, the first one being characterized by convergence, and the other one denoted by divergence. Interestingly, when three historical epochs are analyzed separately (1870-1913; 1913-1950; and 1950-2003), the dynamics which come to dominate over the whole period emerged only during the post-1950 years. In contrast, the First Global Wave was marked by global divergence. Therefore, history does not provide unambiguous evidence about globalization and convergence.

Clustering is unsupervised learning where ideally class levels and number of clusters (K) are not known. Kclustering can be categorized as semi-supervised learning where K is known. Here we have considered K-Clustering with simultaneous feature selection. Feature subset selection helps to identify relevant features for clustering, increase understandability, better scalability and improve accuracy. Here we have used two measures, intra-cluster distance (Homogeneity, H) and inter-cluster distances (Separation, S) for clustering. Measures are using mod distance per feature suitable for categorical features (attributes). Rather than combining H and S to frame the problem as single objective optimization problem, we use multi objective genetic algorithm (MOGA) to find out diverse solutions near to Pareto optimal front in the two-dimensional objective space. Each evolved solution represents a set of cluster modes (CM s) build by selected feature subset. Here, K-modes is hybridized with MOGA. We have used hybridized GA to combine global searching powers of GA with local searching powers of K-modes. Considering context sensitivity, we have used a special crossover operator called "pairwise crossover" and "substitution". The main contribution of this paper is simultaneous dimensionality reduction and optimization of objectives using MOGA. Results on 3 benchmark data sets from UCI Machine Learning Repository containing categorical features shows the superiority of the algorithm.

Data deposition: The new mtDNA sequences were deposited at the EMBL Bank under the accessions HF548556-HF548561. The sequence alignment and the Bayesian consensus tree were deposited at the Dryad Digital Repository (doi:10.5061/dryad.ph920).

The paper proposes Bayesian framework in an M/G/1 queuing system with optional second service. The semi-parametric model based on a finite mixture of Gamma distributions is considered to approximate both the general service and re-service times densities in this queuing system. A Bayesian procedure based on birth-death MCMC methodology is proposed to estimate system parameters, predictive densities and some performance measures related to this queuing system such as stationary system size and waiting time. The approach is illustrated with several numerical examples based on various simulation studies. 123 684 A. Mohammadi et al.

We examine the problem of jointly selecting the number of components and variables in finite mixture regression models. We find that the Akaike information criterion is unsatisfactory for this purpose because it overestimates the number of components, which in turn results in incorrect variables being retained in the model. Therefore, we derive a new information criterion, the mixture regression criterion (MRC), that yields marked improvement in model selection due to what we call the "clustering penalty function." Moreover, we prove the asymptotic efficiency of the MRC. We show that it performs well in Monte Carlo studies for the same or different covariates across components with equal or unequal sample sizes. We also present an empirical example on sales territory management to illustrate the application and efficacy of the MRC. Finally, we generalize the MRC to mixture quasi-likelihood and mixture autoregressive models, thus extending its applicability to non-Gaussian models, discrete responses, and dependent data.

Phylogenetic analyses of DNA sequences were conducted to evaluate four alternative hypotheses of phrynosomatine sand lizard relationships. Sequences comprising 2871 aligned base pair positions representing the regions spanning ND1-COI and cyt b-tRNA Thr of the mitochondrial genome from all recognized sand lizard species were analyzed using unpartitioned parsimony and likelihood methods, likelihood methods with assumed partitions, Bayesian methods with assumed partitions, and Bayesian mixture models. The topology (Uma, (Callisaurus, (Cophosaurus, Holbrookia))) and thus monophyly of the ''earless" taxa, Cophosaurus and Holbrookia, is supported by all analyses. Previously proposed topologies in which Uma and Callisaurus are sister taxa and those in which Holbrookia is the sister group to all other sand lizard taxa are rejected using both parsimony and likelihood-based significance tests with the combined, unparitioned data set. Bayesian hypothesis tests also reject those topologies using six assumed partitioning strategies, and the two partitioning strategies presumably associated with the most powerful tests also reject a third previously proposed topology, in which Callisaurus and Cophosaurus are sister taxa. For both maximum likelihood and Bayesian methods with assumed partitions, those partitions defined by codon position and tRNA stem and nonstems explained the data better than other strategies examined. Bayes factor estimates comparing results of assumed partitions versus mixture models suggest that mixture models perform better than assumed partitions when the latter were not based on functional characteristics of the data, such as codon position and tRNA stem and nonstems. However, assumed partitions performed better than mixture models when functional differences were incorporated. We reiterate the importance of accounting for heterogeneous evolutionary processes in the analysis of complex data sets and emphasize the importance of implementing mixed model likelihood methods.

Most chord recognition systems share a common architecture comprising two main stages: feature extraction and pattern matching, and two optional sub stages: pre-filtering and post-filtering. Understanding the interaction between these basic components is very important not only for achieving optimal performance, but also for assessing the potential and limitations of the system. Unfortunately, there are no studies that sufficiently evaluate the effects of the different approaches to each processing step and the interactions between these steps. In this paper we attempt to remedy this deficiency by performing a systematic evaluation encompassing a wide variety of techniques used for each processing step. In our study we find that filtering has a significant impact on performance, but providing musical context information in the transition matrix is rendered moot by the need to enforce continuity in the estimations. We discovered that the benefits of using complex chord models can be largely offset by an appropriate choice of features. In addition, the initial performance gap between different features were not fully compensated by any subsequent processing stages.

This paper investigates heterogeneity in the preferences/WTP (willingness to pay) to preserve marble monuments in Washington, D.C. This is done in the context of three different discrete-choice random-utility models. The main focus is to estimate a mixture model of choices over preservation programs. This model captures the best features of random-parameters models and models that assume preference parameters are deterministic functions of observable characteristics of the individual. The mixture model, and it alone, predicts that increased preservation is a bad for a significant proportion of young, non-Caucasians. That some proportion of the population might consider preservation a bad is a contingency that should be planned for in efforts to value cultural resources.

This paper investigates recently proposed Stranded Gaussian Mixture acoustic Model (SGMM) for Automatic Speech Recognition (ASR). This model extends conventional hidden Markov model (HMM-GMM) by explicitly introducing dependencies between components of the observation Gaussian mixture densities. The main objective of the paper is to experimentally study, how useful SGMM can be for dealing with data, which contains different sources of acoustic variability. First studied sources of variability are age and gender in quiet environment (TIdigits task including child speech). Second, the SGMM modeling is applied on data produced by different speakers and corrupted by non-stationary noise (CHiME 2013 challenge data). Finally, SGMM is applied on the same noisy data, but after performing speech enhancement (i.e., the remaining variability mostly comes from residual noise and different speakers). Although SGMM was originally proposed for robust speech recognition of noisy data, in this work it was found, that the model is more efficient for handling speaker variability in quiet environment.

We consider a mixture model approach to the regression analysis of competing-risks data. Attention is focused on inference concerning the e ects of factors on both the probability of occurrence and the hazard rate conditional on each of the failure types. These two quantities are speciÿed in the mixture model using the logistic model and the proportional hazards model, respectively. We propose a semi-parametric mixture method to estimate the logistic and regression coe cients jointly, whereby the component-baseline hazard functions are completely unspeciÿed. Estimation is based on maximum likelihood on the basis of the full likelihood, implemented via an expectation-conditional maximization (ECM) algorithm. Simulation studies are performed to compare the performance of the proposed semiparametric method with a fully parametric mixture approach. The results show that when the componentbaseline hazard is monotonic increasing, the semi-parametric and fully parametric mixture approaches are comparable for mildly and moderately censored samples. When the component-baseline hazard is not monotonic increasing, the semi-parametric method consistently provides less biased estimates than a fully parametric approach and is comparable in e ciency in the estimation of the parameters for all levels of censoring. The methods are illustrated using a real data set of prostate cancer patients treated with di erent dosages of the drug diethylstilbestrol.

Mixture model-based methods assuming independence may not be valid for clustering growth trajectories arising from multilevel studies because longitudinal data collected from the same unit are often correlated. A mixture of mixed effects models is considered to capture the correlation using multilevel and multivariate random effects. Furthermore, the mixing proportions are allowed to depend on covariates. The additional information is thus incorporated into the mixture model to adjust for individual probabilities of membership of the components. The proposed method is illustrated using simulated and real multilevel growth trajectory data sets from various scientific fields.

Finite mixture models (FMM) have received increasing attention in recent years and have proven to be useful in modeling heterogeneous data with a finite number of unobserved sub-population. FMM are a powerful and flexible tool for modeling complex data. In this article, we study the asymptotic distribution of the appropriately linear normalized coordinatewise maximum and minimum under mul-tivariate FMM from independent, but not obligatory identically distributed random vectors. We obtain sufficient conditions for this weak convergence, as well as the limit forms. Sufficient conditions for this convergence when the components of the mixture have different linear normalization have been derived. Illustrative examples are provided, which lend further support to our theoretical results.