A new linear initialization in SOM for biomolecular data

In the past decade, the amount of data in biological field has become larger and larger; Bio-techniques for analysis of biological data have been developed and new tools have been introduced. Several computational methods are based on unsupervised neural network algorithms that are widely used for multiple purposes including clustering and visualization, i.e. the Self Organizing Maps (SOM). Unfortunately, even though this method is unsupervised, the performances in terms of quality of result and learning speed are strongly dependent from the neuron weights initialization. In this paper we present a new initialization technique based on a totally connected undirected graph, that report relations among some intersting features of data input. Result of experimental tests, where the proposed algorithm is compared to the original initialization techniques, shows that our technique assures faster learning and better performance in terms of quantization error.

Apprentissage des réseaux de neurones profonds et applications en traitement automatique de la langue naturelle

En apprentissage automatique, domaine qui consiste à utiliser des données pour apprendre une solution aux problèmes que nous voulons confier à la machine, le modèle des Réseaux de Neurones Artificiels (ANN) est un outil précieux. Il a été inventé voilà maintenant près de soixante ans, et pourtant, il est encore de nos jours le sujet d'une recherche active. Récemment, avec l'apprentissage profond, il a en effet permis d'améliorer l'état de l'art dans de nombreux champs d'applications comme la vision par ordinateur, le traitement de la parole et le traitement des langues naturelles. La quantité toujours grandissante de données disponibles et les améliorations du matériel informatique ont permis de faciliter l'apprentissage de modèles à haute capacité comme les ANNs profonds. Cependant, des difficultés inhérentes à l'entraînement de tels modèles, comme les minima locaux, ont encore un impact important. L'apprentissage profond vise donc à trouver des solutions, en régularisant ou en facilitant l'optimisation. Le pré-entraînnement non-supervisé, ou la technique du ``Dropout'', en sont des exemples. Les deux premiers travaux présentés dans cette thèse suivent cette ligne de recherche. Le premier étudie les problèmes de gradients diminuants/explosants dans les architectures profondes. Il montre que des choix simples, comme la fonction d'activation ou l'initialisation des poids du réseaux, ont une grande influence. Nous proposons l'initialisation normalisée pour faciliter l'apprentissage. Le second se focalise sur le choix de la fonction d'activation et présente le rectifieur, ou unité rectificatrice linéaire. Cette étude a été la première à mettre l'accent sur les fonctions d'activations linéaires par morceaux pour les réseaux de neurones profonds en apprentissage supervisé. Aujourd'hui, ce type de fonction d'activation est une composante essentielle des réseaux de neurones profonds. Les deux derniers travaux présentés se concentrent sur les applications des ANNs en traitement des langues naturelles. Le premier aborde le sujet de l'adaptation de domaine pour l'analyse de sentiment, en utilisant des Auto-Encodeurs Débruitants. Celui-ci est encore l'état de l'art de nos jours. Le second traite de l'apprentissage de données multi-relationnelles avec un modèle à base d'énergie, pouvant être utilisé pour la tâche de désambiguation de sens.

Apparent digestibility of diets with different concentrations of lysine and energy in piglets with different body weights and post-weaning age

The effects of body weight or age and dietary digestible lysine and metabolizable energy on apparent digestibility of energy and dry matter were evaluated in piglets after weaning. The animals were weaned at 21 days of age and distributed in two groups: 8.68 ± 0.76 kg at 28 days of age (weaned 7 days earlier); and 12.73 ± 0.99 kg at 35 days of age (weaned 14 days earlier). The pigs were allotted in digestibility cages in a completely randomized block design with the following factorial arrangements: 2 × 4 composed of two weight categories and four levels of digestible lysine (1.222; 1.305; 1.390 and 1.497%); and 2 × 3 composed of two weight categories and three levels of metabolizable energy (3,510; 3,700 and 3,830 kcal/kg rations). Digestible lysine was evaluated in six replications and metabolizable energy in eight replications and each animal constituted an experimental unit. Piglets with higher body weight and age were more efficient in nitrogen retention and energetic balance, compared to lighter and younger piglets, particularly those given lower concentration of lysine in the diet. The energy increase favored nitrogen retention by the heavier and older piglets. However, coefficients of dry matter and energy apparent digestibility did not differ among weight categories. Older and heavier piglets were more efficient in nitrogen retention, although this efficacy depended on concentration of the energy in the diet. This better use of protein and energy suggest differences on nutritional requirements.

Non-additive genetic effects on weights and performance of a Brazilian Bos taurus x Bos indicus beef composite

The aim of the present study was to evaluate the heterosis effects on weaning weight at 205 days (WW, N = 146,464), yearling weight at 390 days (YW, N = 69,315) and weight gain from weaning to yearling (WG, N = 59,307) in composite beef cattle. The fixed models were: RM, which included contemporary groups, class of age of dam, outcrossing percentages for direct and maternal effects, and additive direct and maternal ( AM) breed effects; R, RM model, minus AM breed effects, and H, RM model, minus additive breed effects. The estimates for W205 were in general positive (P < 0.01). The R and H models resulted in similar estimates, but they were very different from the ones estimated by the RM model. For W390, the R and H models resulted in general positive estimates (P < 0.05). For WG, the RM model resulted in general significant heterosis effects (P < 0.05). It can be concluded that the RM model seems to supply estimates of better quality (P < 0.01).

Computed tomographic assessment of lung weights in trauma patients with early posttraumatic lung dysfunction

Introduction: Quantitative computed tomography (qCT)-based assessment of total lung weight (M(lung)) has the potential to differentiate atelectasis from consolidation and could thus provide valuable information for managing trauma patients fulfilling commonly used criteria for acute lung injury (ALI). We hypothesized that qCT would identify atelectasis as a frequent mimic of early posttraumatic ALI. Methods: In this prospective observational study, M(lung) was calculated by qCT in 78 mechanically ventilated trauma patients fulfilling the ALI criteria at admission. A reference interval for M(lung) was derived from 74 trauma patients with morphologically and functionally normal lungs (reference). Results are given as medians with interquartile ranges. Results: The ratio of arterial partial pressure of oxygen to the fraction of inspired oxygen was 560 (506 to 616) mmHg in reference patients and 169 (95 to 240) mmHg in ALI patients. The median reference M(lung) value was 885 (771 to 973) g, and the reference interval for M(lung) was 584 to 1164 g, which matched that of previous reports. Despite the significantly greater median M(lung) value (1088 (862 to 1,342) g) in the ALI group, 46 (59%) ALI patients had M(lung) values within the reference interval and thus most likely had atelectasis. In only 17 patients (22%), Mlung was increased to the range previously reported for ALI patients and compatible with lung consolidation. Statistically significant differences between atelectasis and consolidation patients were found for age, Lung Injury Score, Glasgow Coma Scale score, total lung volume, mass of the nonaerated lung compartment, ventilator-free days and intensive care unit-free days. Conclusions: Atelectasis is a frequent cause of early posttraumatic lung dysfunction. Differentiation between atelectasis and consolidation from other causes of lung damage by using qCT may help to identify patients who could benefit from management strategies such as damage control surgery and lung-protective mechanical ventilation that focus on the prevention of pulmonary complications.

Effects of ambient levels of air pollution generated by traffic on birth and placental weights in mice

Objective: To examine whether there is an association between fetal and/or placental weight and exposure to ambient levels of air pollution in mice. Design: Chronic experiments on mice that were exposed to polluted vs. clean air. Setting: Environmental exposure to atmospheric pollution. Animal(S): Female Swiss mice (n = 70) were maintained at different stages of gestation in an exposure chamber located at an intersection with heavy traffic in a major city in Brazil. Control mice were maintained in a similar chamber, located adjacent to the exposure chamber but equipped with filters for particles and reactive gases. Intervention(s): Animals were divided into six groups as follows: no exposure, exposure to a polluted chamber throughout gestation, exposure to a polluted chamber during the 1st week of pregnancy, exposure to a polluted chamber during the 2nd and 3rd weeks, exposure to a polluted chamber during the 1st and 2nd week, and exposure to a polluted chamber during the 3rd week. Main Outcome Measure(S): At the end of the gestational period, the determination of fetal and placental weight was performed after cesarean section. Result(s): Exposure to air pollution during the 1st week of pregnancy promoted a significant reduction in fetal weight. Mice exposed to polluted air, in any phase of gestation, presented with lower placental weight in comparison to mice maintained in clean chambers. Conclusion(s): Exposure to ambient levels of traffic pollution at early phases of gestation is a determinant for decreased final fetal weight. Placental weight is reduced with exposure to air pollution at any phase of gestation. (Fertil Steril (R) 2008;90:1921-4. (C)2008 by American Society for Reproductive Medicine.)

The Cornerstones of Modern Government Maps, Weights and Measures and Census in Liberal Portugal (19th century)

Thesis submitted for assessment with a view to obtaining the degree of Doctor in History and Civilisation from the European University Institute

On the weights of nations: assigning voting weights in a heterogenous union

Consider a voting procedure where countries, states, or districts comprising a union each elect representatives who then participate in later votes at the union level on their behalf. The countries, provinces, and states may vary in their populations and composition. If we wish to maximize the total expected utility of all agents in the union, how to weight the votes of the representatives of the different countries, states or districts at the union level? We provide a simple characterization of the efficient voting rule in terms of the weights assigned to different districts and the voting threshold (how large a qualified majority is needed to induce change versus the status quo). Next, in the context of a model of the correlation structure of agents preferences, we analyze how voting weights relate to the population size of a country. We then analyze the voting weights in Council of the European Union under the Nice Treaty and the recently proposed constitution, and contrast them under different versions of our model.

Estimation of the Spatial Weights Matrix under Structural Constraints

While estimates of models with spatial interaction are very sensitive to the choice of spatial weights, considerable uncertainty surrounds de nition of spatial weights in most studies with cross-section dependence. We show that, in the spatial error model the spatial weights matrix is only partially identi ed, and is fully identifi ed under the structural constraint of symmetry. For the spatial error model, we propose a new methodology for estimation of spatial weights under the assumption of symmetric spatial weights, with extensions to other important spatial models. The methodology is applied to regional housing markets in the UK, providing an estimated spatial weights matrix that generates several new hypotheses about the economic and socio-cultural drivers of spatial di¤usion in housing demand.

Weights in the balance: jasmonic acid and salicylic acid signaling in root-biotroph interactions.

Work on the interaction of aerial plant parts with pathogens has identified the signaling molecules jasmonic acid (JA) and salicylic acid (SA) as important players in induced defense of the plant against invading organisms. Much less is known about the role of JA and SA signaling in root infection. Recent progress has been made in research on plant interactions with biotrophic mutualists and parasites that exclusively associate with roots, namely arbuscular mycorrhizal and rhizobial symbioses on one hand and nematode and parasitic plant interactions on the other hand. Here, we review these recent advances relating JA and SA signaling to specific stages of root colonization and discuss how both signaling molecules contribute to a balance between compatibility and defense in mutualistic as well as parasitic biotroph-root interactions.

New reference tables and user-friendly Internet application for predicted heart weights.

BACKGROUND: Knowledge of normal heart weight ranges is important information for pathologists. Comparing the measured heart weight to reference values is one of the key elements used to determine if the heart is pathological, as heart weight increases in many cardiac pathologies. The current reference tables are old and in need of an update. AIMS: The purposes of this study are to establish new reference tables for normal heart weights in the local population and to determine the best predictive factor for normal heart weight. We also aim to provide technical support to calculate the predictive normal heart weight. METHODS: The reference values are based on retrospective analysis of adult Caucasian autopsy cases without any obvious pathology that were collected at the University Centre of Legal Medicine in Lausanne from 2007 to 2011. We selected 288 cases. The mean age was 39.2 years. There were 118 men and 170 women. Regression analyses were performed to assess the relationship of heart weight to body weight, body height, body mass index (BMI) and body surface area (BSA). RESULTS: The heart weight increased along with an increase in all the parameters studied. The mean heart weight was greater in men than in women at a similar body weight. BSA was determined to be the best predictor for normal heart weight. New reference tables for predicted heart weights are presented as a web application that enable the comparison of heart weights observed at autopsy with the reference values. CONCLUSIONS: The reference tables for heart weight and other organs should be systematically updated and adapted for the local population. Web access and smartphone applications for the predicted heart weight represent important investigational tools.

Circadian and circaseptan patterns of natality and perinatal mortality of infants with different birth weights

OBJECTIVE. Data on human natality, stillbirth and perinatal mortality from Switzerland (1979-1987), available in four birthweight categories, are reexamined to assess any about-weekly (circaseptan) and changes in about-daily (circadian) patterns in central Europe over a century and a halfDESIGN. Retrospective analyses on archived data.SETTING. Federal Office of Statistics for Switzerland.RESULTS. In addition to prominent circadians, weekly patterns are also documented.CONCLUSION. Exogenous variations, prominent in early extrauterine life, such as changes of scheduling in obstetrics, may contribute to circadian and cireaseptan natality patterns. Information on these patterns serves in the optimization of neonatal care. Partly endogenous, partly physical environmental aspects, at least of about-weekly patterns, remain to be elucidated in series consisting exclusively of spontaneous parturitions.

Spatial weights : constructing weight-compatible exchange matrices from proximity matrices

Exchange matrices represent spatial weights as symmetric probability distributions on pairs of regions, whose margins yield regional weights, generally well-specified and known in most contexts. This contribution proposes a mechanism for constructing exchange matrices, derived from quite general symmetric proximity matrices, in such a way that the margin of the exchange matrix coincides with the regional weights. Exchange matrices generate in turn diffusive squared Euclidean dissimilarities, measuring spatial remoteness between pairs of regions. Unweighted and weighted spatial frameworks are reviewed and compared, regarding in particular their impact on permutation and normal tests of spatial autocorrelation. Applications include tests of spatial autocorrelation with diagonal weights, factorial visualization of the network of regions, multivariate generalizations of Moran's I, as well as "landscape clustering", aimed at creating regional aggregates both spatially contiguous and endowed with similar features.