Éditeur(s) : HAL CCSD Springer Verlag
Résumé : International audience
Based on the very recent work by Censor-Gibali-Reich (http://arxiv.org/abs/1009.3780), we propose an extension of their new variational problem (Split Variational Inequality Problem) to monotone variational inclusions. Relying on the Krasnosel'skii-Mann Theorem for averaged operators, we analyze an algorithm for solving new split monotone inclusions under weaker conditions. Our weak convergence results improve and develop previously discussed Split Variational Inequality Problems, feasibility problems and related problems and algorithms.
ISSN: 0022-3239

hal-00776651
https://hal.univ-antilles.fr/hal-00776651
DOI : 10.1007/s10957-011-9814-6

Éditeur(s) : HAL CCSD Wiley/Blackwell
Résumé : International audience
Microsatellites are widely used in population genetics to uncover recent evolutionary events. They are typically genotyped using capillary sequencer, which capacity is usually limited to 9, at most 12 loci for each run, and which analysis is a tedious task that is performed by hand. With the rise of next-generation sequencing (NGS), a much larger number of loci and individuals are available from sequencing: for example, on a single run of a GS Junior, 28 loci from 96 individuals are sequenced with a 30X cover. We have developed an algorithm to automatically and efficiently genotype microsatellites from a collection of reads sorted by individual (e.g. specific PCR amplifications of a locus or a collection of reads that encompass a locus of interest). As the sequencing and the PCR amplification introduce artefactual insertions or deletions, the set of reads from a single microsatellite allele shows several length variants. The algorithm infers, without alignment, the true unknown allele(s) of each individual from the observed distributions of microsatellites length of all individuals. MicNeSs, a python implementation of the algorithm, can be used to genotype any microsatellite locus from any organism and has been tested on 454 pyrosequencing data of several loci from fruit flies (a model species) and red deers (a nonmodel species). Without any parallelization, it automatically genotypes 22 loci from 441 individuals in 11 hours on a standard computer. The comparison of MicNeSs inferences to the standard method shows an excellent agreement, with some differences illustrating the pros and cons of both methods.
ISSN: 1755-098X

hal-01544754
https://hal.archives-ouvertes.fr/hal-01544754
DOI : 10.1111/1755-0998.12467

Résumé : Les travaux de recherche présentés dans ce mémoire s'inscrivent dans la lignée des approches de modélisation hydrologiques prédictives dirigées par les données. Nous avons particulièrement développé leur application sur le contexte difficile des phénomènes de crue éclairs caractéristiques des bassins versants de la région Caraïbe qui pose un dé fi sé.curi taire. En envisageant le problème de la prévision de crues comme un problème d'optimisation combinatoire difficile nous proposons d'utiliser la notion de métaneuristiques, à travers les algorithmes évolutionnaire notamment pour leur capacité à parcourir efficacement de grands espaces de recherche et fi fournir des solutions de bOlIDe qualité en des temps d'exécution raisonnables. Nous avons présenté l'approche de prédiction AV2D : Aggregate Variable Data Driven dom le concept central est la notion de variable agrégée. L'idée sous-jacente à ce concept est de considérer le pouvoir prédictif de nouvelles variables définies comme le résultat de fonctions tatistiques, dites d'agrégation calculées sur de donnée' correspondant à des périodes de temps précédent uo événem nt à prédire. Ces variable sont caractérisées par des ensembles de paramètres correspondant a leur pJ:opriétés. Nous avons imroduitle variables agrégées hydrométéorologiques permettant de répondre au problème de la classification d événements hydrologiques. La complexité du parcours de l'espace de recherche engendré par les paramètres définissant ces variables a été prise en compte grâce à la njse en oeuvre d'un algorithme évolutionnaire particulier dont les composants ont été spécifiquement définis pour ce problème. Nous avons montré, à travers une étude comparative avec d'autres approches de modélisation dirigées par les données, menée sur deux cas d'études de bassins versant caribéens, que l'approche AV2D est particulièrement bien adaptée à leur contexte. Nous étudions par la suite les bénéfices offerts par les approches de modélisation hydrologiques modulaires dirigées par les données, en définissant un procédé de division en sous-processus prenant en compte les caractéristiques paniculières des bassins versants auxquels nous nous intéressons. Nou avons proposé une extension des travaux précédents à travers la définition d'une approche de modélisation modulaire M2D: Spatial Modular Data Driven, consistant à considérer des sous-processus en divisant l'ensemble des exemples à classifier en sous-ensembles correspondant à des comportements hydrologiques homogènes. Nous avons montré à travers une étude comparative avec d autres approches dU'igées par les données mises en oeuvre sur les mêmes sous-ensembles de données que celte approche permet d améliorer les résultats de prédiction particulièrement à coun Lenne. Nous avons enfin proposé la modélisation d un outil de pi
The work presented in this thesis is in the area of data-driven hydrological modeling approaches. We particularly investigared their application on the difficult problem of flash flood phenomena typically observed in Caribbean watersheds. By considering the problem of flood prediction as a combinatorial optimization problem, we propose to use the notion of Oleraheuristics, through evolutionary algorithms, especially for their capacity ta visit effjciently large search space and to provide good solutions in reasonable execution times. We proposed the hydrological prediction approach AV2D: Aggregate Variable Data Driven which central concept is the notion of aggregate variable. The underlying idea of this [concept is to consider the predictive power of new variables defined as the results of statistical functions, called aggregation functions, computed on data corresponding ta time periods before an event ta predict. These variables are characterized by sets of parameters corresponding ta their specifications. We introduced hydro-meteorological aggregate variables allowing ta address the classification problem of hydrological events. We showed through a comparative study on two typical caribbean watersheds, using several common data driven modelling techniques that the AV2D approach is panicul.rly weil fitted ta the studied context. We also study the benefits offered by modulaI' approaches through the definition of the SM2D: Spatial Modular DataDriven approach, consisting in considering sub-processes partly defined by spatial criteria. We showed that the results obtained by the AV2D on these sub-processes allows to increase the performances particularly for short term prediction. Finally we proposed the modelization of a generic control tool for hydro-meteorological prediction systems, H2FCT: Hydro-meteorological Flood Forecasting Control 1'001

http://www.theses.fr/2013AGUY0673/document

Éditeur(s) : HAL CCSD ACM
Résumé : International audience
This study presents two algorithms that count birds with wireless sensors equipped with microphones. Audio inputs are parametrized to get some kind of fingerprints which are used to recognize the song of the birds in a classification process. Afterward, counting algorithms derive an estimate of the number of singing birds in the habitat. Unlike traditional approaches which leverage the audio waveform to get the location of the targets, we use a trilateration technique on graphs derived from the detection of the birds. These graphs exhibit nice properties which enable the finding of the estimate in polynomial time. Further, experiments are conducted and confirmm the efficiency of our counting algorithms even in the presence of noise.
Proceedings of the 6th ACM symposium on Performance evaluation of wireless ad hoc, sensor, and ubiquitous networks
New York, United States

hal-00602263
https://hal.archives-ouvertes.fr/hal-00602263
DOI : 10.1145/1641876.1641891

Éditeur(s) : HAL CCSD Elsevier
Résumé : International audience
We propose an extended forward-backward algorithm for approximating a zero of a maximal monotone operator which can be split as the extended sum of two maximal monotone operators. We establish the weak convergence in average of the sequence generated by the algorithm under assumptions similar to those used in classical forward-backward algorithms. This provides as a special case an algorithm for solving convex constrained minimization problems without qualification condition.
Journal of Mathematical Analysis and applications

hal-00761657
https://hal.archives-ouvertes.fr/hal-00761657
ARXIV : 1212.0523
DOI : 10.1016/j.jmaa.2013.02.022

Éditeur(s) : HAL CCSD Institute of Electrical and Electronics Engineers
Résumé : International audience
This paper presents our response to the first interna- tional challenge on Facial Emotion Recognition and Analysis. We propose to combine different types of features to automatically detect Action Units in facial images. We use one multi-kernel SVM for each Action Unit we want to detect. The first kernel matrix is computed using Local Gabor Binary Pattern histograms and a histogram intersection kernel. The second kernel matrix is computed from AAM coefficients and an RBF kernel. During the training step, we combine these two types of features using the recently proposed SimpleMKL algorithm. SVM outputs are then averaged to exploit temporal information in the sequence. To eval- uate our system, we perform deep experimentations on several key issues: influence of features and kernel function in histogram- based SVM approaches, influence of spatially-independent in- formation versus geometric local appearance information and benefits of combining both, sensitivity to training data and interest of temporal context adaptation. We also compare our results to those of the other participants and try to explain why our method had the best performance during the FERA challenge.
ISSN: 1083-4419

hal-00731864
http://hal.upmc.fr/hal-00731864
http://hal.upmc.fr/hal-00731864/document
http://hal.upmc.fr/hal-00731864/file/senechal12tsmc_preprint.pdf
DOI : 10.1109/TSMCB.2012.2193567

Éditeur(s) : HAL CCSD
Résumé : International audience
This paper deals with hybrid optimization take different methods to solve a complex problem. As if the choice of concerning methods is important, the way to combine them is determinant for an ef cient search. The article deals with cooperation strategy between the different layers of an hybrid optimization algorithm. We describ the way it is articulated for an Satellite Constellation Optimization problem. This cooperation allows a dynamical adaptation of the optimization algorithm parameters according to the mutual advance of each search
Journal Européen des Systèmes Avancés

hal-00602267
https://hal.archives-ouvertes.fr/hal-00602267

Éditeur(s) : HAL CCSD IOP Publishing
Résumé : International audience
Based on the very recent work by Censor and Segal (2009 J. Convex Anal. 16 587-600) and inspired by Xu (2006 Inverse Problems 22 2021-34) and Yang (2004 Inverse Problems 20 1261-6), we investigate an algorithm for solving the split common fixed-point problem for the class of demicontractive operators in a Hilbert space. Our results improve and/or develop previously discussed feasibility problems and related algorithms. It is worth mentioning that the convex feasibility formalism is at the core of the modeling of many inverse problems and has been used to model significant real-world problems, for instance, in sensor networks, in radiation therapy treatment planning, in computerized tomography and data compression, see Censor et al (2006 Phys. Med. Biol. 51 2353-65) and Combettes (1996 Adv. Imaging Electron. Phys. 95 155-270) and references therein.
ISSN: 0266-5611

hal-00778174
https://hal.univ-antilles.fr/hal-00778174
DOI : 10.1088/0266-5611/26/5/055007

Éditeur(s) : HAL CCSD Lavoisier
Résumé : National audience
Dans cette étude, nous présentons, une méthode permettant de découvrir, par le biais de variables agrégées et d’un algorithme génétique, des informations non connues a priori et utiles pour la prédiction de crues. Cette méthode se prête particulièrement bien aux données dans lesquelles des informations à propos d’un même élément sont dispersées à travers plusieurs lignes du jeu de données initial.
In this paper, we present an heuristic based approach for feature selection in the context of flood prediction. Features are complex variables that represent aggregate values. We apply a preprocessing method on data in order to elicit relevant information that could not be easily accessible initially because it is split through several lines of a dataset. A genetic algorithm is used in order to search for the features that may prove the best performances for flood prediction.
ISSN: 1633-1311

hal-01184510
https://hal.archives-ouvertes.fr/hal-01184510

Éditeur(s) : HAL CCSD Wiley
Résumé : Revue de rang A
International audience
This paper studies how to detect anomalies in a distributed manner by using wireless sensor networks (WSNs). We focus on a medical context, where the existing works generally provide large environments to monitor patients using sensors as simple transducers. Those devices forward sensed health parameters to a main base station. This station collects received data and may perform some computations. In this paper, we perform some distributed tasks on the sensors. We propose a distributed algorithm, which allows to raise alarms under some initial rules to alert efficiently medical staff in case of critical situations without needless warnings. Each mote monitors a parameter. When this parameter reaches an abnormal value, the mote communicates with other motes in order to check if it is a ‘local' anomaly or if the patient is in an abnormal situation. In such cases an alarm is raised. We implemented our algorithm over a network of micaZ sensors running under TinyOS. The obtained results show promising perspectives.
ISSN: 1532-0626

hal-00519594
https://hal.archives-ouvertes.fr/hal-00519594
DOI : 10.1002/cpe.1583

Éditeur(s) : HAL CCSD
Résumé : International audience
A key issue for any project that aims to study endangered animals is the counting process. In this work, we propose an algorithm for automatic counting of singing birds in their habitat by using wireless sensors fitted with microphone. Sensors are used to record audio samples of bird songs and samples are analyzed to extract fingerprints that permit to recognize bird species. This paper details the counting algorithm that uses graph theory and audio inputs comparison. Finally, we demonstrate our scheme efficiency through experimentations.
International Symposium on a World of Wireless, Mobile and Multimedia Networks
Lucca, Italy

hal-00618563
https://hal.archives-ouvertes.fr/hal-00618563
DOI : 10.1109/WoWMoM.2011.5986215

Éditeur(s) : HAL CCSD
Résumé : International audience
This paper provides a simple algorithm for writer identification of historical letters. The collected database is an original historical database, of 100 pages belonging to 25 people selected from a 500 letters database. In the article there is presented an article for a cauterization of the letters, because the system doesn't have the templates of the classes and doesn't know how many classes is in the database. The obtained result shows that automatic identification can help historical experts to segregate the documents, before they would analyze the text information.
17th Biennial Conference of the International Graphonomics Society
Pointe-à-Pitre, Guadeloupe

hal-01165915
https://hal.univ-antilles.fr/hal-01165915
https://hal.univ-antilles.fr/hal-01165915/document
https://hal.univ-antilles.fr/hal-01165915/file/IGS_2015_submission_32.pdf

Éditeur(s) : HAL CCSD Cépaduès-Éditions
Résumé : International audience
Nous présentons dans cet article un nouvel algorithme permettant la construction et la mise à jour incrémentale du FIAθ : FIASCO. Notre algorithme effectue un seul passage sur les données et permet de prendre en compte les nouveaux batches, itemset par itemset et pour chaque itemset, item par item.
EGC: Extraction et Gestion des Connaissances
Nice, France

hal-00447421
https://hal.archives-ouvertes.fr/hal-00447421
https://hal.archives-ouvertes.fr/hal-00447421/document
https://hal.archives-ouvertes.fr/hal-00447421/file/2008--Lionel_Vinceslas--FIASCO_un_nouvel_algorithme_d_extraction_d_itemsets_frequents_dans_les_flots_de_donnees.pdf

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 Oxford University Press (OUP)
Résumé : International audience
OBJECTIVES: The aims of the study were to assess in patients with advanced HIV disease receiving antiretroviral therapy (ART) intensification with enfuvirtide (i) resistance at virological failure (VF), (ii) impact of baseline tropism on immunovirological response, and (iii) HIV-1 DNA tropism evolution during ART. METHODS: The ANRS 130 APOLLO randomized trial evaluated in naive patients the immunovirological impact of standard ART without (control arm) or with enfuvirtide. Tropism was determined on RNA and DNA by V3-loop sequencing interpreted using the Geno2Pheno algorithm. RESULTS: At baseline the median CD4 cell count was 30 cells/mm(3). Among the 170 patients assessable in this virological substudy, HIV-1 RNA tropism was as follows: 60% of viruses were R5 and 40% were R5X4/X4. HIV-1 DNA tropism was as follows: 54% were R5 and 46% were R5X4/X4. At week 24, 39% and 49% of patients experienced VF in the enfuvirtide and control arms, respectively. In the enfuvirtide arm, only resistance-associated mutations to enfuvirtide were detected. In the control arm, two patients displayed drug-resistant viruses at the time of VF. No impact of baseline tropism was observed on immunovirological response, regardless of the study arm. Among the 25 patients experiencing DNA tropism switch between baseline and week 24, 16 (64%) switched from R5 to R5X4/X4. These latter were mostly successfully suppressed patients receiving enfuvirtide and exhibiting poorer immunological response. CONCLUSIONS: Baseline RNA tropism had no impact on the immunovirological response. Drug resistance mutations were only detected for the fusion inhibitor. Finally, the mechanism of replenishment of the viral cellular reservoir with X4 viruses observed needs to be further analysed.
ISSN: 0305-7453

hal-00875130
https://hal.archives-ouvertes.fr/hal-00875130
DOI : 10.1093/jac/dks455
PUBMED : 23152480

Éditeur(s) : HAL CCSD
Résumé : International audience
The aim of this study is to suggest two automated techniques able to help medical staff to detect earlier than usual some diseases using wireless sensor networks (WSNs). In this context, a patient is equipped with physical sensors (for temperature, pressure, ..). This WSN will perform some computations and will run an alarm when a disease is suspected. The first technique uses a population protocol to handle data exchanged between motes and provides an efficient algorithm to suggest that a disease is diagnosed on a patient. The algorithm is distributed, i.e., the decision may be done by any mote dealing with the disease detection. The second technique uses a token algorithm where, some motes, denoted as masters. Each of them is in charge of deciding if a specific disease occurs. This technique is not totally distributed but enhances the network efficiency regarding to the energy consumption, the time execution and the number of exchanged messages.
Proceedings of The Ninth IEEE International Symposium on Networking Computing and Applications
Cambridge, MA, United States

hal-00519599
https://hal.archives-ouvertes.fr/hal-00519599

Éditeur(s) : HAL CCSD School of Communications and Informatics
Résumé : International audience
The hybrid proximal point algorithm introduced by Solodov and Svaiter allowing significant relaxation of the tolerance requirements imposed on the solution of proximal subproblems will be combined with the inertial method introduced by Alvarez and Attouch which incorporates second order information to achieve faster convergence. The weak convergence of the resulting method will be investigated for finding zeroes of a maximal monotone operator in a Hilbert space.
ISSN: 1443-5756

hal-00780745
https://hal.univ-antilles.fr/hal-00780745
https://hal.univ-antilles.fr/hal-00780745/document
https://hal.univ-antilles.fr/hal-00780745/file/170_03_www.pdf

Éditeur(s) : HAL CCSD
Résumé : International audience
Mining sequential patterns in data streams is a new challenging problem for the datamining community since data arrives sequentially in the form of continuous rapid and infinite streams. In this paper, we propose a new on-line algorithm, SPAMS, to deal with the sequential patterns mining problem in data streams. This algorithm uses an automaton-based structure to maintain the set of frequent sequential patterns, i.e. SPA (Sequential Pat- tern Automaton). The sequential pattern automaton can be smaller than the set of frequent sequential patterns by two or more orders of magnitude, which allows us to overcome the problem of combinatorial explosion of se- quential patterns. Current results can be output constantly on any user 's specified thresholds. In addition, taking into account the characteristics of data streams, we propose a well-suited method said to be approximate since we can provide near optimal results with a high probability. Experimental studies show the relevance of the SPA data structure and the efficiency of the SPAMS algorithm on various datasets. Our contribution opens a promis- ing gateway, by using an automaton as a data structure for mining frequent sequential patterns in data streams.
Advances in Knowledge Discovery and Management

lirmm-00435841
https://hal-lirmm.ccsd.cnrs.fr/lirmm-00435841
https://hal-lirmm.ccsd.cnrs.fr/lirmm-00435841/document
https://hal-lirmm.ccsd.cnrs.fr/lirmm-00435841/file/akdmSPAMS.pdf

Éditeur(s) : HAL CCSD Society for Range Management
Résumé : Grasslands and agroecosystems occupy one-third of the terrestrial area, but their contribution to the global carbon cycle remains uncertain. We used a set of 316 site-years of CO(2) exchange measurements to quantify gross primary productivity, respiration, and light-response parameters of grasslands, shrublands/savanna, wetlands, and cropland ecosystems worldwide. We analyzed data from 72 global flux-tower sites partitioned into gross photosynthesis and ecosystem respiration with the use of the light-response method (Gilmanov, T. G., D. A. Johnson, and N. Z. Saliendra. 2003. Growing season CO(2) fluxes in a sagebrushsteppe ecosystem in Idaho: Bowen ratio/energy balance measurements and modeling. Basic and Applied Ecology 4:167-183) from the RANGEFLUX and WORLDGRASSAGRIFLUX data sets supplemented by 46 sites from the FLUXNET La Thuile data set partitioned with the use of the temperature-response method (Reichstein, M., E. Falge, D. Baldocchi, D. Papale, R. Valentini, M. Aubinet, P. Berbigier, C. Bernhofer, N. Buchmann, M. Falk, T. Gilmanov, A. Granier, T. Grunwald, K. Havrankova, D. Janous, A. Knohl, T. Laurela, A. Lohila, D. Loustau, G. Matteucci, T. Meyers, F. Miglietta, J.M. Ourcival, D. Perrin, J. Pumpanen, S. Rambal, E. Rotenberg, M. Sanz, J. Tenhunen, G. Seufert, F. Vaccari, T. Vesala, and D. Yakir. 2005. On the separation of net ecosystem exchange into assimilation and ecosystem respiration: review and improved algorithm. Global Change Biology 11: 1.424-1439). Maximum values of the quantum yield (alpha = 75 mmol.mol(-1)), photosynthetic capacity (A(max) = 3.4 mg CO(2) . m(-2).s-1), gross photosynthesis (P(g,max) = 1.16 g CO(2) . m(-2).d(-1)), and ecological light-use efficiency (epsilon(ecol) = 59 mmol . mol(-1)) of managed grasslands and high-production croplands exceeded those of most forest ecosystems, indicating the potential of nonforest ecosystems for uptake of atmospheric CO(2). Maximum values of gross primary production (8 600 g CO(2) . m(-2).yr(-1)), total ecosystem respiration (7 900 g CO(2) . m(-2).yr(-1)), and net CO(2) exchange (2 400 g CO(2) . m(-2).yr(-1)) were observed for intensively managed grasslands and high-yield crops, and are comparable to or higher than those for forest ecosystems, excluding some tropical forests. On average, 80% of the nonforest sites were apparent sinks for atmospheric CO(2), with mean net uptake of 700 g CO(2) . m(-2).yr(-1) for intensive grasslands and 933 g CO(2) . m(-2).d(-1) for croplands. However, part of these apparent sinks is accumulated in crops and forage, which are carbon pools that are harvested, transported, and decomposed off site. Therefore, although agricultural fields may be predominantly sinks for atmospheric CO(2), this does not imply that they are necessarily increasing their carbon stock.
ISSN: 1550-7424

hal-01190754
https://hal.archives-ouvertes.fr/hal-01190754
DOI : 10.2111/rem-d-09-00072.1
PRODINRA : 214043

Éditeur(s) : HAL CCSD
Résumé : International audience
It is established that the multivalued quasi variational inequalities in uniformly smooth Banach spaces are equivalent to the fixed-point problem. We use this equivalence to suggest and analyze some iterative algorithms for quasi variational inequalities with noncompact sets in Banach spaces. Our results are new and represent a significant improvement of the previously known results.;
JIPAM

hal-00778168
https://hal.univ-antilles.fr/hal-00778168