Éditeur(s) : HAL CCSD IOS Press (Amsterdam)
Résumé : 6 pages, 7 embedded figures, LaTeX 2e using the ecai2006.cls document class and the algorithm2e.sty style file (+ standard packages like epsfig, amsmath, amssymb, amsfonts...). Extends the short version contained in the ECAI 2006 proceedings.
International audience
Without prior knowledge, distinguishing different languages may be a hard task, especially when their borders are permeable. We develop an extension of spectral clustering -- a powerful unsupervised classification toolbox -- that is shown to resolve accurately the task of soft language distinction. At the heart of our approach, we replace the usual hard membership assignment of spectral clustering by a soft, probabilistic assignment, which also presents the advantage to bypass a well-known complexity bottleneck of the method. Furthermore, our approach relies on a novel, convenient construction of a Markov chain out of a corpus. Extensive experiments with a readily available system clearly display the potential of the method, which brings a visually appealing soft distinction of languages that may define altogether a whole corpus.
ECAI 2006: 17th European Conference on Artificial Intelligence
Riva del Garda, Italy

hal-00327782
https://hal.archives-ouvertes.fr/hal-00327782
ARXIV : 0810.1261

Éditeur(s) : HAL CCSD Elsevier
Résumé : International audience
BACKGROUND: Thrombophilias are common disorders that increase the risk of pregnancy-associated venous thromboembolism and pregnancy loss and can also increase the risk of placenta-mediated pregnancy complications (severe pre-eclampsia, small-for-gestational-age infants, and placental abruption). We postulated that antepartum dalteparin would reduce these complications in pregnant women with thrombophilia. METHODS: In this open-label randomised trial undertaken in 36 tertiary care centres in five countries, we enrolled consenting pregnant women with thrombophilia at increased risk of venous thromboembolism or with previous placenta-mediated pregnancy complications. Eligible participants were randomly allocated in a 1:1 ratio to either antepartum prophylactic dose dalteparin (5000 international units once daily up to 20 weeks' gestation, and twice daily thereafter until at least 37 weeks' gestation) or to no antepartum dalteparin (control group). Randomisation was done by a web-based randomisation system, and was stratified by country and gestational age at randomisation day with a permuted block design (block sizes 4 and 8). At randomisation, site pharmacists (or delegates) received a randomisation number and treatment allocation (by fax and/or e-mail) from the central web randomisation system and then dispensed study drug to the local coordinator. Patients and study personnel were not masked to treatment assignment, but the outcome adjudicators were masked. The primary composite outcome was independently adjudicated severe or early-onset pre-eclampsia, small-for-gestational-age infant (birthweight <10th percentile), pregnancy loss, or venous thromboembolism. We did intention-to-treat and on-treatment analyses. This trial is registered with ClinicalTrials.gov, number NCT00967382, and with Current Controlled Trials, number ISRCTN87441504. FINDINGS: Between Feb 28, 2000, and Sept 14, 2012, 292 women consented to participate and were randomly assigned to the two groups. Three women were excluded after randomisation because of ineligibility (two in the antepartum dalteparin group and one in the control group), leaving 146 women assigned to antepartum dalteparin and 143 assigned to no antepartum dalteparin. Some patients crossed over to the other group during treatment, and therefore for on-treatment and safety analysis there were 143 patients in the dalteparin group and 141 in the no dalteparin group. Dalteparin did not reduce the incidence of the primary composite outcome in both intention-to-treat analysis (dalteparin 25/146 [17.1%; 95% CI 11.4-24.2%] vs no dalteparin 27/143 [18.9%; 95% CI 12.8-26.3%]; risk difference -1.8% [95% CI -10.6% to 7.1%)) and on-treatment analysis (dalteparin 28/143 [19.6%] vs no dalteparin 24/141 [17.0%]; risk difference +2.6% [95% CI -6.4 to 11.6%]). In safety analysis, the occurrence of major bleeding did not differ between the two groups. However, minor bleeding was more common in the dalteparin group (28/143 [19.6%]) than in the no dalteparin group (13/141 [9.2%]; risk difference 10.4%, 95% CI 2.3-18.4; p=0.01). INTERPRETATION: Antepartum prophylactic dalteparin does not reduce the occurrence of venous thromboembolism, pregnancy loss, or placenta-mediated pregnancy complications in pregnant women with thrombophilia at high risk of these complications and is associated with an increased risk of minor bleeding. FUNDING: Canadian Institutes of Health Research, Heart and Stroke Foundation of Canada, and Pharmacia and UpJohn.
ISSN: 0923-7577

hal-01258557
http://hal.univ-brest.fr/hal-01258557
DOI : 10.1016/S0140-6736(14)60793-5
PUBMED : 25066248

Éditeur(s) : HAL CCSD
Résumé :  Diffusion du document : INRA Antilles-Guyane, UMR ECOFOG, F97387 Kourou, Guyane (France)Diplôme : Master Recherche
il s'agit d'un type de produit dont les métadonnées ne correspondent pas aux métadonnées attendues dans les autres types de produit : DISSERTATION
Origin and maintenance of species diversity is one of the main questions in evolutive biology. Species complexes, ie sister species still exchanging genes, are common in tropical forests and contribute to this diversity. This poster describes the organization of the genetic diversity in two closely related tree species, Carapa guianensis and C. procera, which are in sympatry in French Guiana. By using a clustering method based on nuclear markers (Pritchard et al. 2000), 506 trees sampled in French Guiana have been assigned to a species. Interestingly, many individuals harbored intermediate genome and we assigned to an hybrid group. The geographic distribution of these individuals revealed the presence of a contact zone between the two carapa species in the Eastern part of French Guiana. We then analysed genetic diversity at cholroplast genome and revealed a symmetrical gene flow between the two sister species. In addition, hybridization between hybrids and parental species ar likely possible but still need validations thanks to control crosses. Within species, we finally revealed the presence of a departure from the neutral equilibrium in thedirection of a population expansion likely related to Holocene perturbations. This expansion may have contributed to the hybrid zone formation in French Guiana.
L'origine et le maintien de la diversité spécifique des forêts tropicales humides est une question clé en biologie évolutive. Ce travail contribue à améliorer la compréhension de ces mécanismes à travers l'étude de la distribution de la diversité génétique de deux espèces sœurs interfertiles, le Carapa procera et C. guianensis en Guyane. L'assignation botanique des arbres échantillonnés (506) a été effectuée en adoptant une approche moléculaire qui repose sur l'analyse de la répartition des fréquences alléliques de six microsatellites nucléaires (Pritchard et al. 2000). Chaque locus présente des allèles dont les différences de fréquences entre espèces sont importantes, permettant ainsi d'envisager des outils simples de reconnaissances pour ces espèces difficilement identifiables sur la seule base de leurs caractères phénotypiques. Une zone de contact des deux espèces où les phénomènes d'hybridations sont fréquents a par ailleurs été localisée. En associant une analyse de la diversité au niveau du génome chloroplastique sur un sous ensemble d'individus (244), j'ai pu montrer que l'introgression est bidirectionnelle et que les backcross sont possibles. Les flux de gènes entre les deux espèces sont fréquents, et à l'intérieur des espèces, les populations sont peu structurées bien que la différenciation soit significative entre les sites les plus éloignés. Enfin, ce travail met en évidence que les populations de chacune des deux espèces présentent un signal d'expansion de population, avec un signal plus fort pour le C. procera, qui laisse présager une expansion récente, vraisemblablement à l'origine de la zone de contact. L'impact des évènements climatiques et/ou des activités amérindiennes des 10 000 dernières années pourrait expliquer la restriction puis l'expansion de ce genre post pionnier.
https://hal.archives-ouvertes.fr/hal-01189488

hal-01189488
https://hal.archives-ouvertes.fr/hal-01189488
PRODINRA : 42333

Éditeur(s) : HAL CCSD Ecological Society of America
Résumé : Reliable assessment of fish origin is of critical importance for exploited species, since nursery areas must be identified and protected to maintain recruitment to the adult stock. During the last two decades, otolith chemical signatures (or "fingerprints") have been increasingly used as tools to discriminate between coastal habitats. However, correct assessment of fish origin from otolith fingerprints depends on various environmental and methodological parameters, including the choice of the statistical method used to assign fish to unknown origin. Among the available methods of classification, Linear Discriminant Analysis (LDA) is the most frequently used, although it assumes data are multivariate normal with homogeneous within-group dispersions, conditions that are not always met by otolith chemical data, even after transformation. Other less constrained classification methods are available, but there is a current lack of comparative analysis in applications to otolith microchemistry. Here, we assessed stock identification accuracy for four classification methods (LDA, Quadratic Discriminant Analysis [QDA], Random Forests [RF], and Artificial Neural Networks [ANN]), through the use of three distinct data sets. In each case, all possible combinations of chemical elements were examined to identify the elements to be used for optimal accuracy in fish assignment to their actual origin. Our study shows that accuracy differs according to the model and the number of elements considered. Best combinations did not include all the elements measured, and it was not possible to define an ad hoc multielement combination for accurate site discrimination. Among all the models tested, RF and ANN performed best, especially for complex data sets (e. g., with numerous fish species and/or chemical elements involved). However, for these data, RF was less time-consuming and more interpretable than ANN, and far more efficient and less demanding in terms of assumptions than LDA or QDA. Therefore, when LDA and QDA assumptions cannot be reached, the use of machine learning methods, such as RF, should be preferred for stock assessment and nursery identification based on otolith microchemistry, especially when data set include multispecific otolith signatures and/or many chemical elements.
ISSN: 1051-0761

hal-00644549
https://hal.archives-ouvertes.fr/hal-00644549
DOI : 10.1890/09-1887.1

Résumé : Ce travail vise à mettre en place un système de fusion de sources d'information basé sur la Théorie des Fonctions de Croyance (TFC) et adapté aux contextes des images forestières. La TFC engendre des temps de calcul importants lorsque la cardinalité du cadre de discernement devient grande. Ainsi, nous avons proposé en premier lieu une méthode de réduction du référentiel de définition permettant d'éviter l'explosion combinatoire en restreignant le calcul sur les hypothèses estimées pertinentes et déterminées à partir des histogrammes des classes issues de la phase d'apprentissage. Ensuite, nous avons proposé des techniques d'estimation de fonctions de masse spectrales, texturales, contextuelles et structurelles. L'information multi-fractale a été utilisée comme signature de texture, ainsi l'image de Hôlder a constitué une source d'information pour l'estimation de la masse texturale. Un autre apport consiste à proposer une approche de transfert de masse basée sur l'information contextuelle, en exploitant l'image des couronnes d'arbres comme une source déduite. L'implémentation d'algorithmes de détection de couronnes d'arbres nous a permis d'extraire des mesures sur les couronnes que nous avons utilisées en tant qu'attributs pour l'estimation des fonctions de masse. Nos expérimentations montrent des taux de classification assez satisfaisants et une adéquation de notre approche de fusion à une classification à l'arbre près
This thesis aims to develop an information source fusion system based on belief functions theory (BFT) and adapted t forest images context (especially high resolution images). Applications of BFT to image classification result in th assignment of each image point to a c1ass or a subset of classes based on the mass functions estimation expressing th belief of an event. BFT leads to significant computation time when the cardinality of the set of definition become large. Thus, we first proposed a method to reduce the frame of discernment in order te avoid the combinatorial explosio by restricting the calculation on the hypotheses of the set of definition deduced from the classes histograms determine from the learning phase. Then, we proposed spectral, texturai, contextual and structural mass functions technique estimation. Multi-fractal information was used as texture signature and Hôlder image has been a source ( information for estimating the texturai mass. Another contribution is the integration of trees crowns image as derive source and the proposai of an approach to mass transfer based on contextual information inside and outside a crown. Th implementation of detection trees crowns algorithms allowed us to extract measurements on trees. These measurement are used as attributes for mass functions estimation. Our experiments and comparisons with other methods show quit satisfactory classification rate with spectral and textural approaches and adequacy of our fusion approach to contextu and structural information for trees classification

http://www.theses.fr/2011AGUY0434/document