Combining opinions for use in Bayesian networks: A measurement error approach

Bayesian networks (BNs) are graphical probabilistic models used for reasoning under uncertainty. These models are becoming increasing popular in a range of fields including ecology, computational biology, medical diagnosis, and forensics. In most of these cases, the BNs are quantified using information from experts, or from user opinions. An interest therefore lies in the way in which multiple opinions can be represented and used in a BN. This paper proposes the use of a measurement error model to combine opinions for use in the quantification of a BN. The multiple opinions are treated as a realisation of measurement error and the model uses the posterior probabilities ascribed to each node in the BN which are computed from the prior information given by each expert. The proposed model addresses the issues associated with current methods of combining opinions such as the absence of a coherent probability model, the lack of the conditional independence structure of the BN being maintained, and the provision of only a point estimate for the consensus. The proposed model is applied an existing Bayesian Network and performed well when compared to existing methods of combining opinions.

Bayesian adaptation and adaptively trained systems

As the use of found data increases, more systems are being built using adaptive training. Here transforms are used to represent unwanted acoustic variability, e.g. speaker and acoustic environment changes, allowing a canonical model that models only the "pure" variability of speech to be trained. Adaptive training may be described within a Bayesian framework. By using complexity control approaches to ensure robust parameter estimates, the standard point estimate adaptive training can be justified within this Bayesian framework. However during recognition there is usually no control over the amount of data available. It is therefore preferable to be able to use a full Bayesian approach to applying transforms during recognition rather than the standard point estimates. This paper discusses various approximations to Bayesian approaches including a new variational Bayes approximation. The application of these approaches to state-of-the-art adaptively trained systems using both CAT and MLLR transforms is then described and evaluated on a large vocabulary speech recognition task. © 2005 IEEE.

End point estimates for Radon transform of radial functions on non-Euclidean spaces

We prove end point estimate for Radon transform of radial functions on affine Grasamannian and real hyperbolic space. We also discuss analogs of these results on the sphere.

Bayesian Accelerated Life Testing under Competing Weibull Causes of Failure

Consider a J-component series system which is put on Accelerated Life Test (ALT) involving K stress variables. First, a general formulation of ALT is provided for log-location-scale family of distributions. A general stress translation function of location parameter of the component log-lifetime distribution is proposed which can accommodate standard ones like Arrhenius, power-rule, log-linear model, etc., as special cases. Later, the component lives are assumed to be independent Weibull random variables with a common shape parameter. A full Bayesian methodology is then developed by letting only the scale parameters of the Weibull component lives depend on the stress variables through the general stress translation function. Priors on all the parameters, namely the stress coefficients and the Weibull shape parameter, are assumed to be log-concave and independent of each other. This assumption is to facilitate Gibbs sampling from the joint posterior. The samples thus generated from the joint posterior is then used to obtain the Bayesian point and interval estimates of the system reliability at usage condition.

Bayesian Accelerated Life Testing under Competing Weibull Causes of Failure

Consider a J-component series system which is put on Accelerated Life Test (ALT) involving K stress variables. First, a general formulation of ALT is provided for log-location-scale family of distributions. A general stress translation function of location parameter of the component log-lifetime distribution is proposed which can accommodate standard ones like Arrhenius, power-rule, log-linear model, etc., as special cases. Later, the component lives are assumed to be independent Weibull random variables with a common shape parameter. A full Bayesian methodology is then developed by letting only the scale parameters of the Weibull component lives depend on the stress variables through the general stress translation function. Priors on all the parameters, namely the stress coefficients and the Weibull shape parameter, are assumed to be log-concave and independent of each other. This assumption is to facilitate Gibbs sampling from the joint posterior. The samples thus generated from the joint posterior is then used to obtain the Bayesian point and interval estimates of the system reliability at usage condition.

A Consumer Surplus Estimate of Peace & Love festival in Borlänge : A Travel Cost Approach

This thesis uses zonal travel cost method (ZTCM) to estimate consumer surplus of Peace & Love festival in Borlänge, Sweden. The study defines counties as zones of origin of the visitors. Visiting rates from each zone are estimated based on survey data. The study is novel due to the fact that mostly TCM has been applied in the environmental and recreational sector, not for short term events, like P&L festival. The analysis shows that travel cost has a significantly negative effect on visiting rate as expected. Even though income has previously shown to be significant in similar studies, it turns out to be insignificant in this study. A point estimate for the total consumer surplus of P&L festival is 35.6 million Swedish kronor. However, this point estimate is associated with high uncertainty since a 95 % confidence interval for it is (17.9, 53.2). It is also important to note that the estimated value only represents one part of the total economic value, the other values of the festival's totaleconomic value have not been estimated in this thesis.

Identification of nonlinear fMRI models using Auxiliary Particle Filter and kernel smoothing method

Hemodynamic models have a high potential in application to understanding the functional differences of the brain. However, full system identification with respect to model fitting to actual functional magnetic resonance imaging (fMRI) data is practically difficult and is still an active area of research. We present a simulation based Bayesian approach for nonlinear model based analysis of the fMRI data. The idea is to do a joint state and parameter estimation within a general filtering framework. One advantage of using Bayesian methods is that they provide a complete description of the posterior distribution, not just a single point estimate. We use an Auxiliary Particle Filter adjoined with a kernel smoothing approach to address this joint estimation problem.

Bayesian Analysis of Linear Inverse Problems with Applications in Economics and Finance

In my PhD thesis I propose a Bayesian nonparametric estimation method for structural econometric models where the functional parameter of interest describes the economic agent's behavior. The structural parameter is characterized as the solution of a functional equation, or by using more technical words, as the solution of an inverse problem that can be either ill-posed or well-posed. From a Bayesian point of view, the parameter of interest is a random function and the solution to the inference problem is the posterior distribution of this parameter. A regular version of the posterior distribution in functional spaces is characterized. However, the infinite dimension of the considered spaces causes a problem of non continuity of the solution and then a problem of inconsistency, from a frequentist point of view, of the posterior distribution (i.e. problem of ill-posedness). The contribution of this essay is to propose new methods to deal with this problem of ill-posedness. The first one consists in adopting a Tikhonov regularization scheme in the construction of the posterior distribution so that I end up with a new object that I call regularized posterior distribution and that I guess it is solution of the inverse problem. The second approach consists in specifying a prior distribution on the parameter of interest of the g-prior type. Then, I detect a class of models for which the prior distribution is able to correct for the ill-posedness also in infinite dimensional problems. I study asymptotic properties of these proposed solutions and I prove that, under some regularity condition satisfied by the true value of the parameter of interest, they are consistent in a "frequentist" sense. Once I have set the general theory, I apply my bayesian nonparametric methodology to different estimation problems. First, I apply this estimator to deconvolution and to hazard rate, density and regression estimation. Then, I consider the estimation of an Instrumental Regression that is useful in micro-econometrics when we have to deal with problems of endogeneity. Finally, I develop an application in finance: I get the bayesian estimator for the equilibrium asset pricing functional by using the Euler equation defined in the Lucas'(1978) tree-type models.

Editoral : Correspondence regarding Anker SD, Koehler F, Abraham WT. Telemedicine and remote management of patients with heart failure. The Lancet 2011; 378: 731–39. Published 20 August 2011.

We read the excellent review of telemonitoring in chronic heart failure (CHF)1 with interest and commend the authors on the proposed classification of telemedical remote management systems according to the type of data transfer, decision ability and level of integration. However, several points require clarification in relation to our Cochrane review of telemonitoring and structured telephone support2. We included a study by Kielblock3. We corresponded directly with this study team specifically to find out whether or not this was a randomised study and were informed that it was a randomised trial, albeit by date of birth. We note in our review2 that this randomisation method carries a high risk of bias. Post-hoc metaanalyses without these data demonstrate no substantial change to the effect estimates for all cause mortality (original risk ratio (RR) 0·66 [95% CI 0·54, 0·81], p<0·0001; revised RR 0·72 [95% CI 0·57, 0·92], p=0·008), all-cause hospitalisation (original RR 0·91 [95% CI 0·84, 0·99] p=0·02; revised RR 0.92 [95% CI 0·84, 1·02], p=0·10 ) or CHF-related hospitalisation (original RR 0·79 [95% CI 0·67, 0·94] p=0·008; revised RR 0·75 [95% CI 0·60, 0·94] p=0·01). Secondly, we would classify the Tele-HF study4, 5 as structured telephone support, rather than telemonitoring. Again, inclusion of these data alters the point-estimate but not the overall result of the meta-analyses4. Finally, our review2 does not include invasive telemonitoring as the search strategy was not designed to capture these studies. Therefore direct comparison of our review findings with recent studies of these interventions is not recommended.

Extended VTS for noise-robust speech recognition

Model compensation is a standard way of improving the robustness of speech recognition systems to noise. A number of popular schemes are based on vector Taylor series (VTS) compensation, which uses a linear approximation to represent the influence of noise on the clean speech. To compensate the dynamic parameters, the continuous time approximation is often used. This approximation uses a point estimate of the gradient, which fails to take into account that dynamic coefficients are a function of a number of consecutive static coefficients. In this paper, the accuracy of dynamic parameter compensation is improved by representing the dynamic features as a linear transformation of a window of static features. A modified version of VTS compensation is applied to the distribution of the window of static features and, importantly, their correlations. These compensated distributions are then transformed to distributions over standard static and dynamic features. With this improved approximation, it is also possible to obtain full-covariance corrupted speech distributions. This addresses the correlation changes that occur in noise. The proposed scheme outperformed the standard VTS scheme by 10% to 20% relative on a range of tasks. © 2006 IEEE.

Estimates of genetic parameters for growth traits of the sea urchin, Strongylocentrotus intermedius

Heritabilities and genetic and phenotypic correlations were estimated for body weight, test diameter, and test height of the sea urchin from measurements on progeny resulting from 11 sires and 33 dams by artificial fertilization of 3 females by single males, and measurements at 8, 10, and 12 months after metamorphism. Point estimate for heritabilities based on the sire components of variance were moderate to high for body weight (0.21-0.49), test diameter (0.21-0.47), and test height (0.22-0.37). Genetic correlations were significant for body weight with test diameter (0.30similar to0.65) and test height (0.30similar to0.54) and test diameter with test height (0.31similar to0.65). Genetic correlation estimates, derived the nested design and half-sib correlation analysis used in this study, appear to provide reliable estimates. Significant phenotypic correlations were found for body weight with test diameter (0.82similar to0.86) and test height (0.49similar to0.83), and test diameter with test height (0.47similar to0.84). The phenotypic correlations for test height with body weight (0.491) and test diameter (0.467) at 12 months' of age were smaller than those earlier sampling periods.(C) 2004 Published by Elsevier B.V.

Estimating the net contribution of interleukin-28B variation to spontaneous hepatitis C virus clearance.

The identification of associations between interleukin-28B (IL-28B) variants and the spontaneous clearance of hepatitis C virus (HCV) raises the issues of causality and the net contribution of host genetics to the trait. To estimate more precisely the net effect of IL-28B genetic variation on HCV clearance, we optimized genotyping and compared the host contributions in multiple- and single-source cohorts to control for viral and demographic effects. The analysis included individuals with chronic or spontaneously cleared HCV infections from a multiple-source cohort (n = 389) and a single-source cohort (n = 71). We performed detailed genotyping in the coding region of IL-28B and searched for copy number variations to identify the genetic variant or haplotype carrying the strongest association with viral clearance. This analysis was used to compare the effects of IL-28B variation in the two cohorts. Haplotypes characterized by carriage of the major alleles at IL-28B single-nucleotide polymorphisms (SNPs) were highly overrepresented in individuals with spontaneous clearance versus those with chronic HCV infections (66.1% versus 38.6%, P = 6 × 10(-9) ). The odds ratios for clearance were 2.1 [95% confidence interval (CI) = 1.6-3.0] and 3.9 (95% CI = 1.5-10.2) in the multiple- and single-source cohorts, respectively. Protective haplotypes were in perfect linkage (r(2) = 1.0) with a nonsynonymous coding variant (rs8103142). Copy number variants were not detected. We identified IL-28B haplotypes highly predictive of spontaneous HCV clearance. The high linkage disequilibrium between IL-28B SNPs indicates that association studies need to be complemented by functional experiments to identify single causal variants. The point estimate for the genetic effect was higher in the single-source cohort, which was used to effectively control for viral diversity, sex, and coinfections and, therefore, offered a precise estimate of the net host genetic contribution.

Mediating Influence of Physical Fitness on the Relationship between Academic Performance and Motor Proficiency

We explored the potential mediating influence of physical fitness on the relationship between academic performance and motor proficiency in children. 1864 students (F:926, M:938, age 11.91 (SD:0.34). Academic achievement was derived from an average of standardized tests of reading, writing, and math. The Bruininks-Oseretsky Test of Motor Performance (short-form) determined motor proficiency. Fitness (peak oxygen uptake) was established with the Léger 20-m Shuttle Run Test. OLS regression identified several significant predictors of academic performance. After controlling for age (p=0.0135), gender (p<0.0001), and parental education (p<0.0001), motor proficiency (p<0.0001), was significant. After adding physical fitness (p=0.0030) to the model the effect of motor proficiency remained significant however the point estimate was reduced from 0.0034 (p<0.0001) to 0.0026 (p<0.0001). These results suggest that physical fitness plays a mediating role on the relationship between academic performance and motor proficiency although both aerobic fitness and motor proficiency have independent roles.

Échantillonnage dynamique de champs markoviens

L'un des modèles d'apprentissage non-supervisé générant le plus de recherche active est la machine de Boltzmann --- en particulier la machine de Boltzmann restreinte, ou RBM. Un aspect important de l'entraînement ainsi que l'exploitation d'un tel modèle est la prise d'échantillons. Deux développements récents, la divergence contrastive persistante rapide (FPCD) et le herding, visent à améliorer cet aspect, se concentrant principalement sur le processus d'apprentissage en tant que tel. Notamment, le herding renonce à obtenir un estimé précis des paramètres de la RBM, définissant plutôt une distribution par un système dynamique guidé par les exemples d'entraînement. Nous généralisons ces idées afin d'obtenir des algorithmes permettant d'exploiter la distribution de probabilités définie par une RBM pré-entraînée, par tirage d'échantillons qui en sont représentatifs, et ce sans que l'ensemble d'entraînement ne soit nécessaire. Nous présentons trois méthodes: la pénalisation d'échantillon (basée sur une intuition théorique) ainsi que la FPCD et le herding utilisant des statistiques constantes pour la phase positive. Ces méthodes définissent des systèmes dynamiques produisant des échantillons ayant les statistiques voulues et nous les évaluons à l'aide d'une méthode d'estimation de densité non-paramétrique. Nous montrons que ces méthodes mixent substantiellement mieux que la méthode conventionnelle, l'échantillonnage de Gibbs.