Éditeur(s) : HAL CCSD Elsevier
Résumé : International audience
Voting rules relying on k-nearest neighbors (k-NN) are an effective tool in countless many machine learning techniques. Thanks to its simplicity, k-NN classification is very attractive to practitioners, as it enables very good performances in several practical applications. However, it suffers from various drawbacks, like sensitivity to "noisy" instances and poor generalization properties when dealing with sparse high-dimensional data. In this paper, we tackle the k-NN classification problem at its core by providing a novel k-NN boosting approach. Namely, we propose a supervised learning algorithm, called Universal Nearest Neighbors (UNN), that induces a leveraged k-NN rule by globally minimizing a surrogate risk upper bounding the empirical misclassification rate over training data. Interestingly, this surrogate risk can be arbitrary chosen from a class of Bregman loss functions, including the familiar exponential, logistic and squared losses. Furthermore, we show that UNN allows to efficiently filter a dataset of instances by keeping only a small fraction of data. Experimental results on the synthetic Ripley's dataset show that such a filtering strategy is able to reject "noisy" examples, and yields a classification error close to the optimal Bayes error. Experiments on standard UCI datasets show significant improvements over the current state of the art.
ISSN: 0925-2312

hal-00664462
https://hal.inria.fr/hal-00664462

Éditeur(s) : Université des Antilles Études caribéennes
Résumé : Si l’île d’Anguilla, dans le nord des Petites Antilles, est une destination secondaire en matière d’arrivées touristiques (moins de 150 000 visiteurs en 2005), il n’en reste pas moins que l’étude des conditions d’émergence du tourisme y est intéressante. Car en la matière, Anguille fait figure d’originalité au sein de son bassin géographique. En revanche, l’analyse des évolutions de la capacité d’hébergement, d’une part, puis des flux de visiteurs, d’autre part, démontre un trait commun à la plupart des îles des Petites Antilles, à savoir une certaine vulnérabilité face à des événements exogènes (catastrophes naturelles et crises géopolitiques). Cet article vise donc à faire le point sur le phénomène touristique (origine, évolution, contraintes) dans une île jusque-là peu étudiée.
Anguilla (91 square kilometres, less than 12 000 inhabitants, 143 200 tourists arrivals in 2005) is a secondary destination in the Lesser Antilles compare to Sint Maarten, Antigua or Barbados. Netherless, tourism development in Anguilla is interesting because it shows that this product is both original (choice of "small and exclusive", majority of excursionnists) and vulnerable (influence of events like cyclone Luis and Lenny in 1995 and 1999, and the terrorist attack of the 2001 september 11th).
Anguilla

Droits : info:eu-repo/semantics/openAccess
urn:doi:10.4000/etudescaribeennes.3492
http://journals.openedition.org/etudescaribeennes/3492

Éditeur(s) : HAL CCSD Elsevier Masson
Résumé :  
Modeling soil water availability for tropical trees is a prerequisite to predicting the future impact of climate change on tropical forests. In this paper we develop a discrete-time deterministic water balance model adapted to tropical rainforest climates, and we validate it on a large dataset that includes micrometeorological and soil parameters along a topographic gradient in a lowland forest of French Guiana. The model computes daily water fluxes (rainfall interception, drainage, tree transpiration and soil plus understorey evapotranspiration) and soil water content using three input variables: daily precipitation, potential evapotranspiration and solar radiation. A novel statistical approach is employed that uses Time Domain Reflectometer (TDR) soil moisture data to estimate water content at permanent wilting point and at field capacity, and root distribution. Inaccuracy of the TDR probes and other sources of uncertainty are taken into account by model calibration through a Bayesian framework. Model daily output includes relative extractable water, REW, i.e. the daily available water standardized by potential available water. The model succeeds in capturing temporal variations in REW regardless of topographic context. The low Root Mean Square Error of Predictions suggests that the model captures the most important drivers of soil water dynamics, i.e. water refilling and root water extraction. Our model thus provides a useful tool to explore the response of tropical forests to climate scenarios of changing rainfall regime and intensity.
ISSN: 0168-1923

hal-01032181
https://hal.archives-ouvertes.fr/hal-01032181
DOI : 10.1016/j.agrformet.2011.04.012

Éditeur(s) : HAL CCSD Wiley
Résumé : International audience
An intensive water vapour intercomparison effort, involving airborne and ground-based water vapour lidar systems, was carried out in the framework of the COPS experiment. The main objective of this paper is to provide accurate error estimates for these systems. Comparisons between the ground-based Raman lidar BASIL and the airborne CNRS DIAL (Differential Absorption Lidar) indicate a mean relative bias between the two sensors, calculated with respect to the mean value of −2.13% (−0.034 g kg−1) in the altitude region 0.5-3.5 km, while comparisons between BASIL and the airborne DLR DIAL lead to a mean relative bias of 1.87% (0.018 g kg−1) in this same altitude region. Comparisons between the ground-based UHOH DIAL and the CNRS DIAL indicate a bias of −3.2% (−0.37 × 1022 m−3) in the altitude range 1.5-4.5 km, while comparisons between the UHOH DIAL and the DLR DIAL indicate a bias of 0.83% (0.06 × 1022 m−3) in this same altitude range. Based on the available comparisons between the ground-based Raman lidar BERTHA and the CNRS DIAL, the mean relative bias is found to be −4.37% (−0.123 g kg−1) in the altitude region 0.5-4.5 km. Comparisons between the ground-based IGN Raman lidar and the CNRS DIAL indicate a bias of 3.18% (0.55 g kg−1) in the altitude range from 0.5 to 4.5 km, while comparisons between the CNRS DIAL and DLR DIAL result in a mean relative bias of 3.93% (1.1 × 1022 m−3) in the altitude interval 0.5-4.0 km. Based on the available statistics of comparisons, benefiting from the fact that the CNRS DIAL was able to be compared with all other lidar systems, and putting equal weight on the data reliability of each instrument, overall relative values for BASIL, BERTHA, IGN Raman lidar, UHOH DIAL, DLR DIAL, and CNRS DIAL, with respect to the mean value, are found to be −0.38, −2.60, 4.90, −1.43, −2.23 and 1.72%, respectively.
ISSN: 0035-9009

hal-00572576
https://hal.archives-ouvertes.fr/hal-00572576
DOI : 10.1002/qj.697

Résumé : (Ownership) The Frank Espada Photographs and Papers were received by the David M. Rubenstein Rare Book & Manuscript Library as a purchase in 2010.
From the Puerto Rican Diaspora Documentary Project
United States
Pennsylvania
Kennett Square (Pa.)

Droits : The copyright in the materials included in the Frank Espada Photographs collection are owned by the photographer. The photographs are made available by Duke University Libraries, with permission, for the purpose of research, teaching, and private study. For these purposes users may reproduce single copies of the images from this website without prior permission, on the condition that proper attribution is provided on all such copies. For all other uses, and especially for any proposed commercial uses, researchers must contact the Library to request permission.
Espada PA 016

Éditeur(s) : EDP Sciences
Résumé : Marine Protected Areas (MPA) are often contemplated as a tool for the sustainable management of exploited resources and ecosystem conservation. This paper proposes an approach to establish a statistical diagnostic of the effects of MPAs on fish assemblages, and define corresponding ecological indicators. This requires choosing relevant variables (abundance, diversity, demographic parameters..) and appropriate statistical methods. The study was based on data from the Abore reef Reserve in New Caledonia. Two sets of methods: 1-inferential linear models (ANOVA, GLM): 2- Partial Least Squares (PLS) methods of regression, were used to test the effects of this MPA. PLS enabled us to test simultaneously within a model, density, species richness, biomass and mean size variables of fish community to retain the most sensitive and relevant ones. Habitat variability was also taken into account in these models. Species were grouped according to several criteria: 1-feeding habit; 2- taxonomy; 3-mobility; 4-adult size; 5-demographic strategy. No significant effect of the opening of the Abore reef to fishing was found for mobility. Feeding habit was the only criterion for which the results from the inferential models and PLS showed a significant effect of reserve status for all variables. Species richness, density, and to a lesser extent mean size, were sensitive to the removal of reserve status. but not biomass. Results from ANOVA and PLS regression were consistent but the latter allows a more holistic approach as it integrates all variables within a single model.
Les Aires Marines Protégées (AMP) sont de plus en plus considérées comme des mesures de gestion alternative permettant à la fois une gestion durable des ressources exploitées et la conservation des écosystèmes. Nous proposons une approche pour définir des indicateurs synthétiques et permettre un diagnostic statistique de l'effet des AMP sur les peuplements. Ce travail passe par le choix de variables pertinentes (abondance, richesse spécifique, paramètres démographiques, etc.) et de méthodes statistiques appropriées. Cette étude a été réalisée sur la réserve du récif Aboré en Nouvelle-Calédonie. Les données issues de deux campagnes scientifiques menées sur ce récif sont utilisées pour tester globalement l'effet des AMP grâce à des modèles linéaires inférentiels (ANOVA, GLM), et par des méthodes de régression PLS (Partial Least Squares). Ces dernières permettent de confronter dans un même modèle quatre variables (densité, richesse, biomasse et taille moyenne des poissons), afin de retenir les plus sensibles et les plus pertinentes pour l'évaluation de l'impact de la réserve. La variabilité de l'habitat est également prise en compte dans la modélisation. Les espèces sont regroupées selon plusieurs critères: 1- le régime alimentaire; 2- la taxonomie; 3- la mobilité; 4- des caractéristiques démographiques; 5- des caractéristiques de taille. Les espèces, quel que soit leur mobilité, n'apparaissent pas affectées par l'ouverture à la pêche du récif Aboré, qu'elles soient sédentaires ou très mobiles. Seul le critère trophique révèle un effet du changement de statut Réserve/Non Réserve pour toutes les variables, à la fois d'après les modèles inférentiels et les régressions PLS. Les résultats des modèles inférentiels et des régressions PLS montrent que la richesse spécifique et la densité sont les variables les plus sensibles à l'impact de l'ouverture à la pêche. La taille moyenne est également bien modélisée sur l'ensemble des résultats mais pas la biomasse. Les résultats d'ANOVA et de régression PLS sont cohérents mais la régression PLS permet d'intégrer toutes les variables dans un même modèle
Aquatic Living Resources (0990-7440) (EDP Sciences), 2004-04 , Vol. 17 , N. 2 , P. 139-149

Droits : EDP Sciences, IFREMER, IRD 2004
http://archimer.ifremer.fr/doc/2004/publication-398.pdf
DOI:10.1051/alr:2004022
http://archimer.ifremer.fr/doc/00000/398/

Éditeur(s) : HAL CCSD Springer Verlag
Résumé : International audience
Fish mucus has important biological and ecologicalroles such as defense against fish pathogens andchemical mediation among several species. A non-targetedliquid chromatography–mass spectrometry metabolomicapproach was developed to study gill mucus of eight butterflyfishspecies in Moorea (French Polynesia), and theinfluence of several fish traits (geographic site and reefhabitat, species taxonomy, phylogeny, diet and parasitismlevels) on the metabolic variability was investigated. Abiphasic extraction yielding two fractions (polar and apolar)was used. Fish diet (obligate corallivorous, facultativecorallivorous or omnivorous) arose as the main driver ofthe metabolic differences in the gill mucus in both fractions,accounting for 23% of the observed metabolic variabilityin the apolar fraction and 13% in the polar fraction.A partial least squares discriminant analysis allowed us toidentify the metabolites (variable important in projection,VIP) driving the differences between fish with differentdiets (obligate corallivores, facultative corallivores andomnivorous). Using accurate mass data and fragmentationdata, we identified some of these VIP as glycerophosphocholines,ceramides and fatty acids. Level of monogeneangill parasites was the second most important factor shapingthe gill mucus metabolome, and it explained 10% of themetabolic variability in the polar fraction and 5% in theapolar fraction. A multiple regression tree revealed that themetabolic variability due to parasitism in the polar fractionwas mainly due to differences between non-parasitized andparasitized fish. Phylogeny and butterflyfish species werefactors contributing significantly to the metabolic variabilityof the apolar fraction (10 and 3%, respectively) buthad a less pronounced effect in the polar fraction. Finally,geographic site and reef habitat of butterflyfish species didnot influence the gill mucus metabolome of butterflyfishes.
ISSN: 0722-4028

hal-01563312
https://hal-univ-perp.archives-ouvertes.fr/hal-01563312
DOI : 10.1007/s00338-017-1554-0

Résumé : Bycatch of marine turtles, vulnerable or endangered species, is a growing issue of all fisheries, including Oceanic purse-seine fishery. The present paper seeks to assess marine turtle bycatch at a spatial and temporal level in the European purse seine fishery operating in the Atlantic and Indian Oceans. The study was based on data collected through French and Spanish observer programs from 1995 to 2011, a period where more than 230 000 fishing sets were realized by the UE fleets in both Oceans. A total of 15 913 fishing sets were observed, including 6 515 on drifting Fish Aggregate Devices (FAD) and 9 398 on Free Swimming Schools (FSC). Over the study period, 597 turtles were caught, 86% being released alive at sea. At the same time, from 2003 to 2011, 14 124 specific observations were carried out on floating objects whether they ended in a set or not. 354 marine turtles were observed upon which 80% were already free or entangled alive and therefore released alive. At the temporal and spatial level, data were organized and analysed by Ocean, fishing mode (FAD vs. FSC) as well as by year, quarter and statistical square of 1°. In order to evaluate the impact if this fishery in both Oceans, bycatch distribution was compared to the total fishing effort of the UE fleet, as well as to the known marine turtle post nesting migration routes, nesting population abundances and known feeding areas. The species composition, the size and sex structure of bycatch are also discussed here. At last, an attempt to raise the data to the total fishing effort was carried out. Based on observation of marine turtle by-catches on sets, we estimated that, globally, 3500 marine turtles were accidentally captured by the EU-PS fleet in the Atlantic Ocean from 1995 to 2010, and around 2000 in the Indian Ocean from 2003 to 2010, with a corresponding annual bycatch rate of 218 (SD=150; survival rate =91%) and 250 (SD=157; survival rate =77%) respectively in the Atlantic and Indian Ocean. However, because of important uncertainties mainly due to the low observation coverage and the scarcity of marine turtle bycatch events, it was impossible to produce solid and reliable global estimates of marine turtle bycatch and mortality due to PS activity.

Droits : 2012 IOTC
http://archimer.ifremer.fr/doc/00149/25991/24087.pdf
http://archimer.ifremer.fr/doc/00149/25991/

Résumé : The slides were taken on collecting trips sponsored by the William L. Bryant Foundation, where books, music and art indigenous to the regions were gathered. The are organized by geographical location.
The Abacos are a group of islands and cays in the Bahamas that form a 120-mile-long chain stretching over 650 square miles. Great Abaco is the second largest island in the Bahamas with one third of the country’s population. Many of smaller islands and cays in the area were settled by British Loyalist after the American Revolutionary War. This image features boats anchored in the harbor of Man of War Cay. On the beach is the Dock and Dine Apartments and Restaurant. Slide labeled Harbor Man of War April '65.
Bahamas -- North America -- Man of War Cay, Abaco

Droits : All rights to images are held by the respective holding institution. This image is posted publicly for non-profit educational uses, excluding printed publication. For permission to reproduce images and/or for copyright information contact Special Collections & University Archives, University of Central Florida Libraries, Orlando, FL 32816 phone (407) 823-2576, email: speccoll@mail.ucf.edu
CFM1972_01a
Sheet 5:11
http://ufdc.ufl.edu/AA00028444/00001

Éditeur(s) : Amer Geophysical Union
Résumé : Sea surface salinity (SSS) measured from space by the Soil Moisture and Ocean Salinity (SMOS) mission is validated in the subtropical North Atlantic Ocean. 39 transects of ships of opportunity equipped with thermosalinographs (TSG) crossed that region from 2010 to 2012, providing a large database of ground truth SSS. SMOS SSS is also compared to Aquarius SSS. Large seasonal biases remain in SMOS and Aquarius SSS. In order to look at the capability of satellite SSS to monitor spatial variability, especially at scales less than 300 km (not monitored with the Argo network), we first apply a monthly bias correction derived from satellite SSS and In Situ Analysis System (ISAS) SSS differences averaged over the studied region. Ship SSS averaged over 25 km is compared with satellite and ISAS SSS. Similar statistics are obtained for SMOS, Aquarius and ISAS products (root mean square error of about 0.15 and global correlation coefficient r of about 0.92). However, in the above statistics, SSS varies due to both large scale and mesoscale (here, for scales around 100 km) variability. In order to focus on mesoscale variability, we consider SSS anomalies with respect to a monthly climatology. SMOS SSS and Aquarius SSS anomalies are more significantly correlated (r > 0.5) to TSG SSS anomaly than ISAS. We show the effective gain of resolution and coverage provided by the satellite products over the interpolated in situ data. We also show the advantage of SMOS (r=0.57) over Aquarius (r=0.52) to reproduce SSS mesoscale features.
Journal Of Geophysical Research-oceans (0148-0027) (Amer Geophysical Union), 2014-12 , Vol. 119 , N. 12 , P. 8878-8896

Droits : 2014. American Geophysical Union. All Rights Reserved.
http://archimer.ifremer.fr/doc/00210/32150/34082.pdf
DOI:10.1002/2013JC009610
http://archimer.ifremer.fr/doc/00210/32150/

Éditeur(s) : HAL CCSD
Résumé : A standard approach for large scale image classification involves high dimensional features and Stochastic Gradient Descent algorithm (SGD) for the minimization of classical Hinge Loss in the primal space. Although complexity of Stochastic Gradient Descent is linear with the number of samples these method suffers from slow convergence. In order to cope with this issue, we propose here a Stochastic Low-Rank Newton Descent SLND for minimization of any calibrated loss in the primal space. SLND approximates the inverse Hessian by the best low-rank approximation according to squared Frobenius norm. We provide core optimization for fast convergence. Theoretically speaking, we show explicit convergence rates of the algorithm using these calibrated losses, which in addition provide working sets of parameters for experiments. Experiments are provided on the SUN, Caltech256 and ImageNet databases, with simple, uniform and efficient ways to tune remaining SLND parameters. On each of these databases, SLND challenges the accuracy of SGD with a speed of convergence faster by order of magnitude.
https://hal.archives-ouvertes.fr/hal-00825414

hal-00825414
https://hal.archives-ouvertes.fr/hal-00825414
https://hal.archives-ouvertes.fr/hal-00825414/document
https://hal.archives-ouvertes.fr/hal-00825414/file/TechnicalReport.pdf

Éditeur(s) : HAL CCSD Wiley
Résumé : International audience
The Convective and Orographically-driven Precipitation Study (COPS) carried out in summer 2007 over northeastern France and southwestern Germany provided a fairly comprehensive description of the low-troposphere water-vapour field, thanks in particular to the deployment of two airborne differential absorption lidar systems. These lidar observations were assimilated using the 3D-Var assimilation system of the Application of Research to Operations at MEsoscale (AROME) numerical weather prediction mesoscale model. The assimilation was carried out for the period 4 July-3 August by running a three-hour forward intermittent assimilation cycle. First, the impact of the lidar observations was assessed by comparing the analyses with a set of more than 200 independent soundings. The lidar observations were found to have a positive impact on the analyses by reducing the dry bias in the first 500 m above ground level and by diminishing the root-mean-square error by roughly 15% in the first km. Then the impact of the lidar observations was assessed by comparing the precipitation forecasts (obtained with and without the lidar observations for the period 15 July-2 August) with the gridded precipitation observations provided by the Vienna Enhanced Resolution Analysis. In general, the impact was found to be positive but not significant for the 24 h precipitation and positive and significant for the 6 h precipitation, with an improvement lasting up to 24 h. Some selected case studies show that the improvement was obtained through a better depiction of convection initiation or through a more accurate positioning of the precipitation systems
ISSN: 0035-9009

hal-00659497
https://hal.archives-ouvertes.fr/hal-00659497
DOI : 10.1002/qj.1864

Éditeur(s) : HAL CCSD Public Library of Science
Résumé : International audience
BACKGROUND: The use of pesticides and the related environmental contaminations can lead to human exposure to various molecules. In early-life, such exposures could be responsible for adverse developmental effects. However, human health risks associated with exposure to complex mixtures are currently under-explored. OBJECTIVE: THIS PROJECT AIMS AT ANSWERING THE FOLLOWING QUESTIONS: What is the influence of exposures to multiple pesticides on the metabolome? What mechanistic pathways could be involved in the metabolic changes observed? METHODS: Based on the PELAGIE cohort (Brittany, France), 83 pregnant women who provided a urine sample in early pregnancy, were classified in 3 groups according to the surface of land dedicated to agricultural cereal activities in their town of residence. Nuclear magnetic resonance-based metabolomics analyses were performed on urine samples. Partial Least Squares Regression-Discriminant Analysis (PLS-DA) and polytomous regressions were used to separate the urinary metabolic profiles from the 3 exposure groups after adjusting for potential confounders. RESULTS: The 3 groups of exposure were correctly separated with a PLS-DA model after implementing an orthogonal signal correction with pareto standardizations (R2 = 90.7% and Q2 = 0.53). After adjusting for maternal age, parity, body mass index and smoking habits, the most statistically significant changes were observed for glycine, threonine, lactate and glycerophosphocholine (upward trend), and for citrate (downward trend). CONCLUSION: This work suggests that an exposure to complex pesticide mixtures induces modifications of metabolic fingerprints. It can be hypothesized from identified discriminating metabolites that the pesticide mixtures could increase oxidative stress and disturb energy metabolism.
ISSN: 1932-6203

inserm-00863407
http://www.hal.inserm.fr/inserm-00863407
http://www.hal.inserm.fr/inserm-00863407/document
http://www.hal.inserm.fr/inserm-00863407/file/journal.pone.0064433.pdf
PUBMED : 23704985

Éditeur(s) : HAL CCSD Elsevier
Résumé : International audience
The correlation between the flow turbulence and the performances of a ma- rine current turbine is studied. First, the incoming flow encountered in the flume tank is characterized in the framework of fully developed turbulent cascades in the inertial range. The Reynolds number, the Kolmogorov dissipation scale and the integral scale, are estimated from flow measurements. The intermittency of the turbulence is characterized in the lognormal multifractal framework, and the influence of the turbulent flow on the turbine power is assessed. The rotor speed control unit characteristics used for the turbine regulation induces non-negligible effects on the turbine behavior under fluctuations loads. Even if the power spectrum does not reveal any scale invariance, a multiscale analysis allows us to show the correlations between the turbulence time se- ries and the power produced. The classical Mean Square Coherency function shows that for scales larger than 10 seconds, the upstream velocity and power have large correlations. In the framework of the Empirical Mode Decomposition method, such correlations are studied using the time-dependence intrinsic corre- lation analysis method. This method allows to zoom into time-frequency scales where the flow perturbations induced some modifications in power production
ISSN: 0960-1481

hal-01528296
https://hal.archives-ouvertes.fr/hal-01528296
DOI : 10.1016/j.renene.2017.05.024

Éditeur(s) : Elsevier Sci Ltd
Résumé : A multi-scale global hindcast of ocean waves is presented that covers the years 1994-2012, based on recently published parameterizations for wind sea and swell dissipation [Ardhuin, F., Rogers, E., Babanin, A., Filipot, J.-F., Magne, R., Roland, A., van der Westhuysen, A., Queffeulou, P., Lefevre, J.-M., Aouf, L., Collard, F., 2010. Semi-empirical dissipation source functions for wind-wave models: Part I. Definition, calibration and validation. J. Phys. Oceanogr. 40 (9), 1917-1941]. Results from this hindcast include traditional wave parameters, like the significant wave height and mean periods, and we particularly consider the accuracy of the results for phenomenal sea states, with significant heights above 14 m. Using unbiased winds, there is no evidence of a bias in wave heights even for this very high range. Various spectral moments were also validated, including the surface Stokes drift and mean square slopes that are relevant for wave-current interactions modelling and remote sensing, and also spectra of seismic noise sources. The estimation of these parameters is made more accurate by the new wave growth and dissipation parameterizations. Associated air-sea fluxes of momentum and energy are significantly different from what is obtained with the WAM-Cycle 4 parameterization, with a roughness that is practically a function of wind speed only. That particular output of the model does not appear very realistic and will require future adjustments of the generation and dissipation parameterizations.
Ocean Modelling (1463-5003) (Elsevier Sci Ltd), 2013-10 , Vol. 70 , P. 174-188

Droits : 2012 Elsevier Ltd. All rights reserved.
http://archimer.ifremer.fr/doc/00155/26582/25477.pdf
DOI:10.1016/j.ocemod.2012.12.001
http://archimer.ifremer.fr/doc/00155/26582/

Éditeur(s) : HAL CCSD
Résumé : International audience
ABSTRACT :A satellite retrieval of surface solar irradiance (SSI) based on GOES satellite data is presented and validated for the French Guiana with three years of in situ measurements from six ground stations. We propose a particle filter approach combining GOES satellite observations and in situ data. We propose an original observation function based on a joint probability distribution and taking advantages of the characteristics of both the involved type of data. The statistical results are compared with those of existing estimation methods and they are found to be in accordance with them. The daily bias ranges from -12 W/m² (-6% of the mean of measurements) to 12 W/m² (6%), depending on the stations. The daily root mean square difference ranges between 23 W/m² (10%) and 27 W/m² (15%). The correlation coefficient is close to 0.92. There is no seasonal climatic effect on the method, the bias and RMSD remain similar for both the rainy and the dry season, but better correlation is observed in the rainy season than in the dry season. Uncertainties are mainly due to the presence of specific tropical atmospheric conditions during the dry season. The method has been tested through various tests: cross-validation scheme has highlighted the potential generalization of the model in a temporal dimension; the ability of the method to effectively duplicate anomalies in surface solar irradiance time series has also been tested. The robustness of the results shows the interest of the method compared to other existing estimation methods. Because of its simplicity of implementation this method opens new prospects in estimating surface solar irradiance. It is concluded that using a particle filter and original observation function method allow a high-quality surface solar irradiance estimation in French Guiana.RESUMEUne méthode d'estimation de l'irradiance solaire de surface (SSI) à partir de données satellites GOES est présentée et validée par comparaison avec trois années de mesures in situ provenant de six stations météorologiques de Guyane Française. Dans cet article, nous proposons une approche par filtre particulaire combinant observations du satellite GOES et données in situ. Nous proposons une fonction d'observation originale basée sur une distribution de probabilité conjointe des deux types de données. Les résultats statistiques sont conformes à ceux des méthodes existantes. Le biais journalier varie de -16 W / m² (-7% de la moyenne des mesures) à 12 W / m² (6%). L'erreur quadratique moyenne (RMSD) est comprise entre 23 W / m² (10%) et 27 W / m² (15%). Le coefficient de corrélation est de 0,92. La méthode est peu sensible aux variations climatiques saisonnières, le biais et RMSD restent similaires en saison des pluies comme en saison sèche, mais une meilleure corrélation est observée pendant la saison des pluies. Les incertitudes sont principalement dues à la présence de conditions atmosphériques tropicales spécifiques en saison sèche. La méthode a été soumise à différents tests : un processus de validation croisée a mis en évidence la possible généralisation du modèle dans une dimension temporelle, la capacité du modèle à reproduire efficacement des anomalies dans les séries temporelles d'irradiance solaire de surface a aussi été testée. La robustesse des résultats obtenus montrent l'intérêt du modèle par rapport aux autres méthodes d'estimation existantes. A cause de sa simplicité de mise en oeuvre cette méthode ouvre de nouvelles perspectives en matière d'estimation de l'irradiance solaire au sol. Nous concluons que l'utilisation d'un filtre particulaire combinée à une fonction d'observation originale permet une estimation de haute qualité de l'irradiance solaire au sol en Guyane Française.
Canadian Journal of Remote Sensing

meteo-01304574
https://hal-meteofrance.archives-ouvertes.fr/meteo-01304574
DOI : 10.1080/07038992.2015.1040150

Éditeur(s) : HAL CCSD ACM New York, NY, USA
Résumé : International audience
This paper focuses on the study of the behavior of a genetic algorithm based classifier system, the Adapted Pittsburgh Classifier System (A.P.C.S), on maze type environments containing aliasing squares. This type of environment is often used in reinforcement learning litterature to assess the performances of learning methods when facing problems containing non markov situations. Through this study, we discuss on the performances of the APCS on maze type environments and also of the efficiency of an improvement of the APCS learning method inspired from the XCS : the covering mechanism. We manage to show that, without any memory mechanism, the APCS is able to build and to keep accurate strategies to produce regular sub-optimal solutions to these maze problem. This statement is shown through a comparison of the results obtained by the XCS, XCSM and XCSMH on distinct maze problems with these obtained by the APCS.
Proceedings of the 2008 GECCO conference companion on Genetic and evolutionary computation
Atlanta, GA, United States

hal-00520607
https://hal.archives-ouvertes.fr/hal-00520607
DOI : 10.1145/1388969.1389013

Éditeur(s) : HAL CCSD MDPI
Résumé : International audience
Satellite precipitation products are a means of estimating rainfall, particularly in areas that are sparsely equipped with rain gauges. The Guiana Shield is a region vulnerable to high water episodes. Flood risk is enhanced by the concentration of population living along the main rivers. A good understanding of the regional hydro-climatic regime, as well as an accurate estimation of precipitation is therefore of great importance. Unfortunately, there are very few rain gauges available in the region. The objective of the study is then to compare satellite rainfall estimation products in order to complement the information available in situ and to perform a regional analysis of four operational precipitation estimates, by partitioning the whole area under study into a homogeneous hydro-climatic region. In this study, four satellite products have been tested, TRMM TMPA (Tropical Rainfall Measuring Mission Multisatellite Precipitation Analysis) V7 (Version 7) and RT (real time), CMORPH (Climate Prediction Center (CPC) MORPHing technique) and PERSIANN (Precipitation Estimation from Remotely-Sensed Information using Artificial Neural Network), for daily rain gauge data. Product performance is evaluated at daily and monthly scales based on various intensities and hydro-climatic regimes from 1 January 2001 to 30 December 2012 and using quantitative statistical criteria (coefficient correlation, bias, relative bias and root mean square error) and quantitative error metrics (probability of detection for rainy days and for no-rain days and the false alarm ratio). Over the entire study period, all products underestimate precipitation. The results obtained in terms of the hydro-climate show that for areas with intense convective precipitation, TMPA V7 shows a better performance than other products, especially in the estimation of extreme precipitation events. In regions along the Amazon, the use of PERSIANN is better. Finally, in the driest areas, TMPA V7 and PERSIANN show the same performance.
ISSN: 2072-4292

Droits : http://hal.archives-ouvertes.fr/licences/copyright/
hal-01257912
https://hal.archives-ouvertes.fr/hal-01257912
https://hal.archives-ouvertes.fr/hal-01257912/document
https://hal.archives-ouvertes.fr/hal-01257912/file/remotesensing-07-15831.pdf
DOI : 10.3390/rs71215831

Éditeur(s) : HAL CCSD Elsevier
Résumé : Displacement-length data from faults, joints, veins, igneous dikes, shear deformation bands, and compaction bands define two groups. The first group, having a power-law scaling relation with a slope of n = 1 and therefore a linear dependence of maximum displacement and discontinuity length (D-max = gamma L), comprises faults and shear (non-compactional or non-dilational) deformation bands. These shearing-mode structures, having shearing strains that predominate over volumetric strains across them, grow under conditions of constant driving stress, with the magnitude of near-tip stress on the same order as the rock's yield strength in shear. The second group, having a power-law scaling relation with a slope of n = 0.5 and therefore a dependence of maximum displacement on the square root of discontinuity length (D-max = alpha L-0.5), comprises joints, veins, igneous dikes, cataclastic deformation bands, and compaction bands. These opening- and closing-mode structures grow under conditions of constant fracture toughness, implying significant amplification of near-tip stress within a zone of small-scale yielding at the discontinuity tip. Volumetric changes accommodated by grain fragmentation, and thus control of propagation by the rock's fracture toughness, are associated with scaling of predominantly dilational and compactional structures with an exponent of n = 0.5.
ISSN: 0191-8141

hal-00411173
https://hal.archives-ouvertes.fr/hal-00411173
DOI : 10.1016/j.jsg.2008.08.001

Éditeur(s) : HAL CCSD
Résumé : International audience
MOTIVATION: Imaging mass spectrometry (IMS) is a maturating technique of molecular imaging. Confidence in the reproducible quality of IMS data is essential for its integration into routine use. However, the predominant method for assessing quality is visual examination, a time consuming, unstandardized and non-scalable approach. So far, the problem of assessing the quality has only been marginally addressed and existing measures do not account for the spatial information of IMS data. Importantly, no approach exists for unbiased evaluation of potential quality measures. RESULTS: We propose a novel approach for evaluating potential measures by creating a gold-standard set using collective expert judgements upon which we evaluated image-based measures. To produce a gold standard, we engaged 80 IMS experts, each to rate the relative quality between 52 pairs of ion images from MALDI-TOF IMS datasets of rat brain coronal sections. Experts' optional feedback on their expertise, the task and the survey showed that (i) they had diverse backgrounds and sufficient expertise, (ii) the task was properly understood, and (iii) the survey was comprehensible. A moderate inter-rater agreement was achieved with Krippendorff's alpha of 0.5. A gold-standard set of 634 pairs of images with accompanying ratings was constructed and showed a high agreement of 0.85. Eight families of potential measures with a range of parameters and statistical descriptors, giving 143 in total, were evaluated. Both signal-to-noise and spatial chaos-based measures performed highly with a correlation of 0.7 to 0.9 with the gold standard ratings. Moreover, we showed that a composite measure with the linear coefficients (trained on the gold standard with regularized least squares optimization and lasso) showed a strong linear correlation of 0.94 and an accuracy of 0.98 in predicting which image in a pair was of higher quality. AVAILABILITY AND IMPLEMENTATION: The anonymized data collected from the survey and the Matlab source code for data processing can be found at: https://github.com/alexandrovteam/IMS_quality. CONTACT: theodore.alexandrov@embl.de
Bioinformatics (Oxford, England)

hal-01163754
https://hal-univ-rennes1.archives-ouvertes.fr/hal-01163754
DOI : 10.1093/bioinformatics/btv266
PUBMED : 26072506