Stratégie de vaccination familiale contre la coqueluche (méthode de cocooning) à la maternité : analyse coût-efficacité d'un programme provincial

Le manuscrit constituant l'annexe 1 a été publié en décembre 2013 sous la référence : Vaccine. 2013 Dec 9;31(51):6087-91.

Modeling High-Dimensional Audio Sequences with Recurrent Neural Networks

Cette thèse étudie des modèles de séquences de haute dimension basés sur des réseaux de neurones récurrents (RNN) et leur application à la musique et à la parole. Bien qu'en principe les RNN puissent représenter les dépendances à long terme et la dynamique temporelle complexe propres aux séquences d'intérêt comme la vidéo, l'audio et la langue naturelle, ceux-ci n'ont pas été utilisés à leur plein potentiel depuis leur introduction par Rumelhart et al. (1986a) en raison de la difficulté de les entraîner efficacement par descente de gradient. Récemment, l'application fructueuse de l'optimisation Hessian-free et d'autres techniques d'entraînement avancées ont entraîné la recrudescence de leur utilisation dans plusieurs systèmes de l'état de l'art. Le travail de cette thèse prend part à ce développement. L'idée centrale consiste à exploiter la flexibilité des RNN pour apprendre une description probabiliste de séquences de symboles, c'est-à-dire une information de haut niveau associée aux signaux observés, qui en retour pourra servir d'à priori pour améliorer la précision de la recherche d'information. Par exemple, en modélisant l'évolution de groupes de notes dans la musique polyphonique, d'accords dans une progression harmonique, de phonèmes dans un énoncé oral ou encore de sources individuelles dans un mélange audio, nous pouvons améliorer significativement les méthodes de transcription polyphonique, de reconnaissance d'accords, de reconnaissance de la parole et de séparation de sources audio respectivement. L'application pratique de nos modèles à ces tâches est détaillée dans les quatre derniers articles présentés dans cette thèse. Dans le premier article, nous remplaçons la couche de sortie d'un RNN par des machines de Boltzmann restreintes conditionnelles pour décrire des distributions de sortie multimodales beaucoup plus riches. Dans le deuxième article, nous évaluons et proposons des méthodes avancées pour entraîner les RNN. Dans les quatre derniers articles, nous examinons différentes façons de combiner nos modèles symboliques à des réseaux profonds et à la factorisation matricielle non-négative, notamment par des produits d'experts, des architectures entrée/sortie et des cadres génératifs généralisant les modèles de Markov cachés. Nous proposons et analysons également des méthodes d'inférence efficaces pour ces modèles, telles la recherche vorace chronologique, la recherche en faisceau à haute dimension, la recherche en faisceau élagué et la descente de gradient. Finalement, nous abordons les questions de l'étiquette biaisée, du maître imposant, du lissage temporel, de la régularisation et du pré-entraînement.

Realisation of a Target Classifier For Noise Sources in the Ocean

This thesis addresses one of the emerging topics in Sonar Signal Processing.,viz.the implementation of a target classifier for the noise sources in the ocean, as the operator assisted classification turns out to be tedious,laborious and time consuming.In the work reported in this thesis,various judiciously chosen components of the feature vector are used for realizing the newly proposed Hierarchical Target Trimming Model.The performance of the proposed classifier has been compared with the Euclidean distance and Fuzzy K-Nearest Neighbour Model classifiers and is found to have better success rates.The procedures for generating the Target Feature Record or the Feature vector from the spectral,cepstral and bispectral features have also been suggested.The Feature vector ,so generated from the noise data waveform is compared with the feature vectors available in the knowledge base and the most matching pattern is identified,for the purpose of target classification.In an attempt to improve the success rate of the Feature Vector based classifier,the proposed system has been augmented with the HMM based Classifier.Institutions where both the classifier decisions disagree,a contention resolving mechanism built around the DUET algorithm has been suggested.

Online Learning of Non-stationary Sequences

We consider an online learning scenario in which the learner can make predictions on the basis of a fixed set of experts. The performance of each expert may change over time in a manner unknown to the learner. We formulate a class of universal learning algorithms for this problem by expressing them as simple Bayesian algorithms operating on models analogous to Hidden Markov Models (HMMs). We derive a new performance bound for such algorithms which is considerably simpler than existing bounds. The bound provides the basis for learning the rate at which the identity of the optimal expert switches over time. We find an analytic expression for the a priori resolution at which we need to learn the rate parameter. We extend our scalar switching-rate result to models of the switching-rate that are governed by a matrix of parameters, i.e. arbitrary homogeneous HMMs. We apply and examine our algorithm in the context of the problem of energy management in wireless networks. We analyze the new results in the framework of Information Theory.

Diseño de una vía clínica para la vigilancia de hepatocarcinoma y evaluación económica

El carcinoma Hepatocelular (HCC) representa la sexta causa más frecuente de cáncer, y la tercera causa de muerte relacionada con cáncer en el mundo con aproximadamente 600.000 muertes anuales. En el 70 % de los casos, este se desarrolla en presencia de una enfermedad crónica del hígado como la cirrosis u otras enfermedades inflamatorias, por lo que practicar métodos de tamizaje para su diagnóstico precoz, pudieran establecer un mejor pronóstico. El objetivo de este trabajo es diseñar una vía clínica capaz de homogenizar el proceso de tamizaje de HCC, soportando su realización con la realización de una evaluación económica de esta intervención. Se realiza una búsqueda sistemática de literatura y se propone una vía clínica para la vigilancia de HCC en Colombia. A esta propuesta se aplica una evaluación económica tipo costo-efectividad mediante un modelo de Markov de la intervención propuesta, comparando la aplicación de la vía clínica propuesta frente al manejo actual en 100 pacientes considerados con riesgo (cirrosis, portadores de HBV y/o portadores de HCV) con un horizonte de tiempo de 30 años analizando como desenlace los años de vida salvados (LYS) desde la perspectiva del tercero pagador para Colombia a precios de 2009. El análisis determina una disminución de la mortalidad en un 40%, y un valor ICER de US$ 1,438 por LYS, por lo cual se concluye que resulta costo efectivo la aplicación de esta propuesta de tamizaje. Es necesario realizar una prueba para la aplicación de la vía clínica.

Estudio Costo-Efectividad de Lidocaína tópica como adyuvante a Pregabalina y Gabapentina en dolor neuropático periférico en una I.P.S de Bogotá 2011.

Introducción: el dolor neuropático es una patología de considerable prevalencia e impacto socio-económico en la población latinoamericana, la evidencia clínica sugiere que los ligandos de canales de calcio y el parche de Lidocaína pueden tratar exitosamente el dolor neuropático periférico y localizado. Metodología: se realizo una evaluación económica tipo costo-efectividad, observacional y retrospectiva con datos extraídos de las historias clínicas de pacientes atendidos en la clínica de dolor de la IPS. La variable primaria de efectividad fue la mejoría del dolor medida mediante escala visual análoga. Resultados: se estudiaron 94 pacientes tratados con: Gabapentina (G) 21, Pregabalina (P) 24, Gabapentina+ lidocaína (G/P) 24, Pregabalina + Lidocaína (P/L) 25, los costos asociados al tratamiento son los siguientes COP$114.070.835, COP$105.855.920, COP$88.717.481 COP$89.854.712 respectivamente, el número de pacientes con mejoría significativa de dolor fue: 8,10,9 y 21 pacientes respectivamente. El ICER de G/L con respecto a G fue: COP$ -25.353.354. El ICER de P/L con respecto a P fue: COP$ -1.454.655. Conclusiones: la adición del parche de lidocaína a la terapia regular con P/L represento una disminución de consumo de recursos en salud como uso de medicamentos co-analgésicos, analgésicos de rescate y fármacos para controlar reacciones adversas, de la misma forma que consultas a profesionales de la salud. Cada paciente manejado con P/L representa un ahorro de COP $1.454.655 al contrario si se manejase con el anticonvulsivante de manera exclusiva, en el caso de G/L este ahorro es de COP $ 25.353.354 frente a G sola.

Characterizing PvARP, a novel Plasmodium vivax antigen

Background Plasmodium vivax continues to be the most widely distributed malarial parasite species in tropical and sub-tropical areas, causing high morbidity indices around the world. Better understanding of the proteins used by the parasite during the invasion of red blood cells is required to obtain an effective vaccine against this disease. This study describes characterizing the P. vivax asparagine-rich protein (PvARP) and examines its antigenicity in natural infection. Methods The target gene in the study was selected according to a previous in silico analysis using profile hidden Markov models which identified P. vivax proteins that play a possible role in invasion. Transcription of the arp gene in the P. vivax VCG-1 strain was here evaluated by RT-PCR. Specific human antibodies against PvARP were used to confirm protein expression by Western blot as well as its subcellular localization by immunofluorescence. Recognition of recombinant PvARP by sera from P. vivax-infected individuals was evaluated by ELISA. Results VCG-1 strain PvARP is a 281-residue-long molecule, which is encoded by a single exon and has an N-terminal secretion signal, as well as a tandem repeat region. This protein is expressed in mature schizonts and is located on the surface of merozoites, having an apparent accumulation towards their apical pole. Sera from P. vivax-infected patients recognized the recombinant, thereby suggesting that this protein is targeted by the immune response during infection.

Can the frequency of blocking be described by a red noise process?

The frequency of persistent atmospheric blocking events in the 40-yr ECMWF Re-Analysis (ERA-40) is compared with the blocking frequency produced by a simple first-order Markov model designed to predict the time evolution of a blocking index [defined by the meridional contrast of potential temperature on the 2-PVU surface (1 PVU ≡ 1 × 10−6 K m2 kg−1 s−1)]. With the observed spatial coherence built into the model, it is able to reproduce the main regions of blocking occurrence and the frequencies of sector blocking very well. This underlines the importance of the climatological background flow in determining the locations of high blocking occurrence as being the regions where the mean midlatitude meridional potential vorticity (PV) gradient is weak. However, when only persistent blocking episodes are considered, the model is unable to simulate the observed frequencies. It is proposed that this persistence beyond that given by a red noise model is due to the self-sustaining nature of the blocking phenomenon.

Aggregation of parasitism risk in an aphid-parasitoid system: Effects of plant patch size and aphid density

Few studies have linked density dependence of parasitism and the tritrophic environment within which a parasitoid forages. In the non-crop plant-aphid, Centaurea nigra-Uroleucon jaceae system, mixed patterns of density-dependent parasitism by the parasitoids Aphidius funebris and Trioxys centaureae were observed in a survey of a natural population. Breakdown of density-dependent parasitism revealed that density dependence was inverse in smaller colonies but direct in large colonies (>20 aphids), suggesting there is a threshold effect in parasitoid response to aphid density. The CV2 of searching parasitoids was estimated from parasitism data using a hierarchical generalized linear model, and CV2>1 for A. funebris between plant patches, while for T. centaureae CV2>1 within plant patches. In both cases, density independent heterogeneity was more important than density-dependent heterogeneity in parasitism. Parasitism by T. centaureae increased with increasing plant patch size. Manipulation of aphid colony size and plant patch size revealed that parasitism by A. funebris was directly density dependent at the range of colony sizes tested (50-200 initial aphids), and had a strong positive relationship with plant patch size. The effects of plant patch size detected for both species indicate that the tritrophic environment provides a source of host density independent heterogeneity in parasitism, and can modify density-dependent responses. (c) 2007 Gessellschaft fur Okologie. Published by Elsevier GmbH. All rights reserved.

Analysis of the calving pattern of herds in small-scale dairy systems in central Mexico

This study sets out to find the best calving pattern for small-scale dairy systems in Michoacan State, central Mexico. Two models were built. First, a linear programming model was constructed to optimize calving pattern and herd structure according to metabolizable energy availability. Second, a Markov chain model was built to investigate three reproductive scenarios (good, average and poor) in order to suggest factors that maintain the calving pattern given by the linear programming model. Though it was not possible to maintain the optimal linear programming pattern, the Markov chain model suggested adopting different reproduction strategies according to period of the year that the cow is expected to calve. Comparing different scenarios, the Markov model indicated the effect of calving interval on calving pattern and herd structure.

A combined Bayesian Markovian approach for behaviour recognition

Numerous techniques exist which can be used for the task of behavioural analysis and recognition. Common amongst these are Bayesian networks and Hidden Markov Models. Although these techniques are extremely powerful and well developed, both have important limitations. By fusing these techniques together to form Bayes-Markov chains, the advantages of both techniques can be preserved, while reducing their limitations. The Bayes-Markov technique forms the basis of a common, flexible framework for supplementing Markov chains with additional features. This results in improved user output, and aids in the rapid development of flexible and efficient behaviour recognition systems.

Brain computer interface control via functional connectivity dynamics

The dynamics of inter-regional communication within the brain during cognitive processing – referred to as functional connectivity – are investigated as a control feature for a brain computer interface. EMDPL is used to map phase synchronization levels between all channel pair combinations in the EEG. This results in complex networks of channel connectivity at all time–frequency locations. The mean clustering coefficient is then used as a descriptive feature encapsulating information about inter-channel connectivity. Hidden Markov models are applied to characterize and classify dynamics of the resulting complex networks. Highly accurate levels of classification are achieved when this technique is applied to classify EEG recorded during real and imagined single finger taps. These results are compared to traditional features used in the classification of a finger tap BCI demonstrating that functional connectivity dynamics provide additional information and improved BCI control accuracies.

Dynamic game soundtrack generation in response to a continously varying emotional trajectory

Dynamic soundtracking presents various practical and aesthetic challenges to composers working with games. This paper presents an implementation of a system addressing some of these challenges with an affectively-driven music generation algorithm based on a second order Markov-model. The system can respond in real-time to emotional trajectories derived from 2-dimensions of affect on the circumplex model (arousal and valence), which are mapped to five musical parameters. A transition matrix is employed to vary the generated output in continuous response to the affective state intended by the gameplay.

Stochastic simulations of the Madden-Julian oscillation activity

The Madden-Julian oscillation (MJO) is the most prominent form of tropical intraseasonal variability. This study investigated the following questions. Do inter-annual-to-decadal variations in tropical sea surface temperature (SST) lead to substantial changes in MJO activity? Was there a change in the MJO in the 1970s? Can this change be associated to SST anomalies? What was the level of MJO activity in the pre-reanalysis era? These questions were investigated with a stochastic model of the MJO. Reanalysis data (1948-2008) were used to develop a nine-state first order Markov model capable to simulate the non-stationarity of the MJO. The model is driven by observed SST anomalies and a large ensemble of simulations was performed to infer the activity of the MJO in the instrumental period (1880-2008). The model is capable to reproduce the activity of the MJO during the reanalysis period. The simulations indicate that the MJO exhibited a regime of near normal activity in 1948-1972 (3.4 events year(-1)) and two regimes of high activity in 1973-1989 (3.9 events) and 1990-2008 (4.6 events). Stochastic simulations indicate decadal shifts with near normal levels in 1880-1895 (3.4 events), low activity in 1896 1917 (2.6 events) and a return to near normal levels during 1918-1947 (3.3 events). The results also point out to significant decadal changes in probabilities of very active years (5 or more MJO events): 0.214 (1880-1895), 0.076 (1896-1917), 0.197 (1918-1947) and 0.193 (1948-1972). After a change in behavior in the 1970s, this probability has increased to 0.329 (1973-1989) and 0.510 (1990-2008). The observational and stochastic simulations presented here call attention to the need to further understand the variability of the MJO on a wide range of time scales.

Semantic mapping using mobile robots

Robotic mapping is the process of automatically constructing an environment representation using mobile robots. We address the problem of semantic mapping, which consists of using mobile robots to create maps that represent not only metric occupancy but also other properties of the environment. Specifically, we develop techniques to build maps that represent activity and navigability of the environment. Our approach to semantic mapping is to combine machine learning techniques with standard mapping algorithms. Supervised learning methods are used to automatically associate properties of space to the desired classification patterns. We present two methods, the first based on hidden Markov models and the second on support vector machines. Both approaches have been tested and experimentally validated in two problem domains: terrain mapping and activity-based mapping.