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

Résumé : Pour la formalisation du lexique et de la grammaire de dialectes étroitement apparentés, il peut se révéler utile de factoriser une partie du travail de modélisation. Les soussystèmes linguistiques isomorphes dans les différents dialectes peuvent alors faire l’objet d’une description commune, les différences étant spécifiées par ailleurs. Cette démarche aboutit à un modèle de grammaire à couches : le noyau est commun à la famille de dialectes, et une couche superficielle détermine les caractéristiques de chacun. Nous appliquons ce procédé à la famille des langues créoles à base lexicale française de l’aire américano-caraïbe.
The task of writing formal lexicons and grammars for closely related dialects can benefit from factoring part of the modelling. Isomorphic linguistic subsystems from the different dialectsmay have a common description, while the differences are specified aside. This process leads to a layered grammar model: a kernel common to the whole family of dialects, and a superficial skin specifying the particular properties of each one of them. We apply this principle to the family of French-lexifier creole languages of the American-Caribbean area.

Droits : Ce document est protégé par le droit d'auteur. Il ne peut en aucun cas être utilisé sans l'autorisation de l'auteur et des ayant droits
recherch:HASH0177c3f56b04eebb1ecda993

Résumé : Pour la formalisation du lexique et de la grammaire de dialectes étroitement apparentés, il peut se révéler utile de factoriser une partie du travail de modélisation. Les soussystèmes linguistiques isomorphes dans les différents dialectes peuvent alors faire l’objet d’une description commune, les différences étant spécifiées par ailleurs. Cette démarche aboutit à un modèle de grammaire à couches : le noyau est commun à la famille de dialectes, et une couche superficielle détermine les caractéristiques de chacun. Nous appliquons ce procédé à la famille des langues créoles à base lexicale française de l’aire américano-caraïbe.
The task of writing formal lexicons and grammars for closely related dialects can benefit from factoring part of the modelling. Isomorphic linguistic subsystems from the different dialectsmay have a common description, while the differences are specified aside. This process leads to a layered grammar model: a kernel common to the whole family of dialects, and a superficial skin specifying the particular properties of each one of them. We apply this principle to the family of French-lexifier creole languages of the American-Caribbean area.

Droits : Ce document est protégé par le droit d'auteur. Il ne peut en aucun cas être utilisé sans l'autorisation de l'auteur et des ayant droits
recherch:HASH0177c3f56b04eebb1ecda993

Éditeur(s) : HAL CCSD Cambridge University Press (CUP)
Résumé : International audience
We extend the theory of distributional kernel operators to a framework of generalized functions, in which they are replaced by integral kernel operators. Moreover, in contrast to the distributional case, we show that these generalized integral operators can be composed unrestrictedly. This leads to the definition of the exponential, and more generally entire functions, of a subclass of such operators.
ISSN: 0305-0041

hal-01530755
https://hal.univ-antilles.fr/hal-01530755

Éditeur(s) : HAL CCSD Wiley
Résumé : International audience
In this paper we investigate finite-frequency effects in crustal tomography. We developed an inversion procedure based on an exact numerical computation of the sensitivity kernels. In this approach we compute the 3D travel-time sensitivity kernels by using (1) graph theory and an additional bending to estimate accurately both rays and travel-times between source/receiver and diffraction points and (2) paraxial ray theory to estimate the amplitude along theses rays. We invert both the velocity and the hypocentre parameters, using these so-called banana-doughnut kernels and the LSQR iterative solver. We compare the ray-theoretical and the finite-frequency tomography to image the intermediate structures beneath the Gulf of Corinth (Greece), which has long been recognized as the most active continental rifting zone in the Mediterranean region. Our dataset consists of 451 local events with 9233 P- firstarrival times recorded in the western part of the Gulf (Aigion area) in the framework of the 3F-Corinth European project. Previous tomographic images showed a complex velocity crustal model and a low-dip surface that may accommodate the deformation. Accurate velocity models will help to better constrain the rifting process, which is still a subject of debate. The main results of this study show that finite-frequency tomography improves crustal tomographic images by providing better resolved images of the 3D complicated velocity structure. Because the kernels spread the information over a volume, finite-frequency tomography results in a sharpening of layer boundaries as we observed for the shallower part of the crust (down to 5 km depth) beneath the Gulf of Corinth.
ISSN: 0016-8025

insu-00354702
https://hal-insu.archives-ouvertes.fr/insu-00354702
DOI : 10.1111/j.1365-2478.2007.00683.x

Éditeur(s) : HAL CCSD
Résumé : International audience
The k-nearest neighbors (k-NN) classification rule is still an essential tool for computer vision applications, such as scene recognition. However, k-NN still features some major drawbacks, which mainly reside in the uniform voting among the nearest prototypes in the feature space. In this paper, we propose a new method that is able to learn the "relevance" of prototypes, thus classifying test data using a weighted k-NN rule. In particular, our algorithm, called Multi-class Leveraged k-nearest neighbor (MLNN), learns the prototype weights in a boosting framework, by minimizing a surrogate exponential risk over training data. We propose two main contributions for improving computational speed and accuracy. On the one hand, we implement learning in an inherently multiclass way, thus providing significant computation time reduction over one-versus-all approaches. Furthermore, the leveraging weights enable effective data selection, thus reducing the cost of k-NN search at classification time. On the other hand, we propose a kernel generalization of our approach to take into account real-valued similarities between data in the feature space, thus enabling more accurate estimation of the local class density. We tested MLNN on three datasets of natural images. Results show that MLNN significantly outperforms classic k-NN and weighted k-NN voting. Furthermore, using an adaptive Gaussian kernel provides significant performance improvement. Finally, the best results are obtained when using MLNN with an appropriate learned metric distance.
Proceedings of the 10th Asian Conference on Computer Vision, ACCV 2010, November 8-12, 2010, Queenstown, New Zealand
Queenstown, New Zealand

hal-00664606
https://hal.inria.fr/hal-00664606
https://hal.inria.fr/hal-00664606/document
https://hal.inria.fr/hal-00664606/file/mlnn_accvfinal.pdf

Éditeur(s) : HAL CCSD Elsevier
Résumé : International audience
In analogy to the classical Schwartz kernel theorem, we show that a large class of linear mappings admits integral kernels in the framework of Colombeau generalized functions. To do this, we introduce new spaces of generalized functions with slow growth and the corresponding adapted linear mappings. Finally, we show that, in some sense, Schwartz’ result is contained in our main theorem.
ISSN: 0022-247X

hal-01530748
https://hal.univ-antilles.fr/hal-01530748
https://hal.univ-antilles.fr/hal-01530748/document
https://hal.univ-antilles.fr/hal-01530748/file/delcroixgskttjan21.pdf
DOI : 10.1016/j.jmaa.2005.01.006

Éditeur(s) : HAL CCSD
Résumé : 25 pages, Equipe AANL
We extend the theory of distributional kernel operators to a framework of generalized functions, in which they are replaced by integral kernel operators. Moreover, in contrast to the distributional case, we show that these generalized integral operators can be composed unrestrictedly. This leads to the definition of the exponential, and more generally entire functions, of a subclass of such operators.
https://hal.archives-ouvertes.fr/hal-00004888

hal-00004888
https://hal.archives-ouvertes.fr/hal-00004888
https://hal.archives-ouvertes.fr/hal-00004888/document
https://hal.archives-ouvertes.fr/hal-00004888/file/Article5.pdf
ARXIV : math/0505126

Éditeur(s) : HAL CCSD
Résumé : To appear in the Proceeding of GF2004 in ITSF (8 pages).
We extend the theory of distributional kernel operators to a framework of generalized functions, in which they are replaced by integral kernel operators. Moreover, in contrast to the distributional case, we show that these generalized integral operators can be composed unrestrictedly. This leads to the definition of the exponential of a subclass of such operators.
https://hal.archives-ouvertes.fr/hal-00004897

hal-00004897
https://hal.archives-ouvertes.fr/hal-00004897
https://hal.archives-ouvertes.fr/hal-00004897/document
https://hal.archives-ouvertes.fr/hal-00004897/file/GF04BernardDelcroixColombeauNov01.pdf
ARXIV : math.GM/0505179

Éditeur(s) : HAL CCSD
Résumé : International audience
Colloque en l'honneur de Robert Lutz.
Mulhouse, France

hal-01535440
https://hal.univ-antilles.fr/hal-01535440

Éditeur(s) : Public Library Science
Résumé : Studying the larval dispersal of bottom-dwelling species is necessary to understand their population dynamics and optimize their management. The black-lip pearl oyster (Pinctada margaritifera) is cultured extensively to produce black pearls, especially in French Polynesia's atoll lagoons. This aquaculture relies on spat collection, a process that can be optimized by understanding which factors influence larval dispersal. Here, we investigate the sensitivity of P. margaritifera larval dispersal kernel to both physical and biological factors in the lagoon of Ahe atoll. Specifically, using a validated 3D larval dispersal model, the variability of lagoon-scale connectivity is investigated against wind forcing, depth and location of larval release, destination location, vertical swimming behavior and pelagic larval duration (PLD) factors. The potential connectivity was spatially weighted according to both the natural and cultivated broodstock densities to provide a realistic view of connectivity. We found that the mean pattern of potential connectivity was driven by the southwest and northeast main barotropic circulation structures, with high retention levels in both. Destination locations, spawning sites and PLD were the main drivers of potential connectivity, explaining respectively 26%, 59% and 5% of the variance. Differences between potential and realistic connectivity showed the significant contribution of the pearl oyster broodstock location to its own dynamics. Realistic connectivity showed larger larval supply in the western destination locations, which are preferentially used by farmers for spat collection. In addition, larval supply in the same sectors was enhanced during summer wind conditions. These results provide new cues to understanding the dynamics of bottom-dwelling populations in atoll lagoons, and show how to take advantage of numerical models for pearl oyster management.
Plos One (1932-6203) (Public Library Science), 2014-04 , Vol. 9 , N. 4 , P. e95050

Droits : 2014 Thomas et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
http://archimer.ifremer.fr/doc/00197/30843/29210.pdf
DOI:10.1371/journal.pone.0095050
http://archimer.ifremer.fr/doc/00197/30843/

Éditeur(s) : HAL CCSD Taylor & Francis
Résumé : International audience
We extend the theory of distributional kernel operators to a framework of generalized functions, in which they are replaced by integral kernel operators. Moreover, in contrast to the distributional case, we show that these generalized integral operators can be composed unrestrictedly. This leads to the definition of the exponential, and more generally entire functions, of a subclass of such operators.
ISSN: 1065-2469

hal-01530753
https://hal.univ-antilles.fr/hal-01530753

Description : Pour la formalisation du lexique et de la grammaire de dialectes étroitement apparentés, il peut se révéler utile de factoriser une partie du travail de modélisation. Les soussystèmes linguistiques isomorphes dans les différents dialectes peuvent alors faire l'objet d'une description commune, les différences étant spécifiées par ailleurs. Cette démarche aboutit à un modèle de grammaire à couches : le noyau est commun à la famille de dialectes, et une couche superficielle détermine les caractéristiques de chacun. Nous appliquons ce procédé à la famille des langues créoles à base lexicale française de l'aire américano-caraïbe.

Droits : Ce document est protégé par le droit d'auteur. Il ne peut en aucun cas être utilisé sans l'autorisation de l'auteur et des ayant droits
Provenance : Université des Antilles et de la Guyane. Service commun de la documentation
Permalien : http://www.manioc.org/recherch/HASH0177c3f56b04eebb1ecda993
HASH0177c3f56b04eebb1ecda993

Éditeur(s) : HAL CCSD Institute of Mathematics Polish Academy of Sciences
Résumé : International audience
In analogy to the classical isomorphism between L(D(Rn);D0(Rm)) and D0(Rm+n) (resp. L(S(Rn); S0(Rm)) and S0(Rm+n)), we show that a large class of moderate linear mappings acting between the space GC(Rn) of compactly supported generalized functions and G(Rn) of generalized functions (resp. the space GS(Rn) of Colombeau rapidly decreasing generalized functions and the space G (Rn) of temperate ones) admits generalized integral representations, with kernels belonging to specic regular subspaces of G(Rm+n) (resp. G (Rm+n)). The main novelty is to use accelerated -nets, which are unit elements for the convolution product in these regular subspaces, to construct the kernels. Finally, we establish a strong relationship between these results and the classical ones.
ISSN: 0137-6934

hal-01529541
https://hal.univ-antilles.fr/hal-01529541
https://hal.univ-antilles.fr/hal-01529541/document
https://hal.univ-antilles.fr/hal-01529541/file/delcroix.pdf
DOI : 10.4064/bc88-0-7

Éditeur(s) : HAL CCSD
Résumé : International audience
Generalized functions 2004
Novi Sad, Serbia

hal-01535462
https://hal.univ-antilles.fr/hal-01535462

Éditeur(s) : HAL CCSD Institute of Electrical and Electronics Engineers
Résumé : International audience
The state-space modeling of partially observed dynamical systems generally requires estimates of unknown parameters. The dynamic state vector together with the static parameter vector can be considered as an augmented state vector. Classical filtering methods, such as the extended Kalman filter (EKF) and the bootstrap particle filter (PF), fail to estimate the augmented state vector. For these classical filters to handle the augmented state vector, a dynamic noise term should be artificially added to the parameter components or to the deterministic component of the dynamical system. However, this approach degrades the estimation performance of the filters. We propose a variant of the PF based on convolution kernel approximation techniques. This approach is tested on a simulated case study.
ISSN: 0018-9251

inria-00459886
https://hal.inria.fr/inria-00459886
DOI : 10.1109/TAES.2009.5259183

Éditeur(s) : Elsevier
Résumé : Overexploitation of marine resources remains a problem worldwide. Many works advocate for the use of marine reserves as a central element of future stock management in a sustainable perspective. In the present paper, we address the influence of protected areas upon fisheries sustainability within an eco-systemic framework through a dynamic bio-economic model integrating a trophic web, catches and environmental uncertainties. The model is spatially implicit. The evaluation of the ecosystem is designed through the respect along time of constraints of both conservation and guaranteed captures. Using the mathematical concept of invariance kernel in a stochastic context, we define different MPA effects according to biodiversity catches or mixed points of view. Numerical simulations inspired from data of Abore coral reef reserve in New Caledonia illustrate the main concepts. In this case, it is pointed out how MPA conservation effect is not necessarily conflicting with MPA catches effect. it is shown that such a co-viability requires medium exploitation rate. Moreover, the climatic changes represented by rise in cyclonic events seem to reinforce these assertions. (C) 2007 Elsevier B.V. All rights reserved.
Ecological Modelling (0304-3800) (Elsevier), 2007-11 , Vol. 208 , N. 2-4 , P. 353-366

Droits : 2007 Elsevier B.V. All rights reserved.
http://archimer.ifremer.fr/doc/2007/publication-3510.pdf
DOI:10.1016/j.ecolmodel.2007.06.018
http://archimer.ifremer.fr/doc/00000/3510/

Éditeur(s) : HAL CCSD
Résumé : Session "Articles"
National audience
Dans cet article, nous présentons une méthode robuste et précise pour détecter 17 points caractéristiques du visage sur des images expressives. Une nouvelle architecture multi-résolution basée sur les récents algorithmes multi-noyau est introduite. Les patches de faibles résolutions codent les informations globales du visage donnant lieu à une détection grossière mais robuste du point désiré. Les patches de grandes résolutions quant à eux utilisent les détails locaux afin d'affiner cette localisation. En combinant une détection indépendante de points et des informations a priori sur les distributions de points, nous proposons un détecteur robuste aux changements d'expressions ainsi qu'aux variations d'illuminations. Ce système a été évalué sur plusieurs bases de données de la littérature. Les résultats présentés améliorent les performances des détecteurs de l'état de l'art
Actes de la conférence RFIA 2012
Lyon, France

hal-00656522
https://hal.archives-ouvertes.fr/hal-00656522
https://hal.archives-ouvertes.fr/hal-00656522/document
https://hal.archives-ouvertes.fr/hal-00656522/file/rfia2012_submission_99.pdf

Éditeur(s) : HAL CCSD
Résumé : 15 pages
International audience
In analogy to the classical isomorphism between $mathcal{L}left( mathcal{S}left( mathbb{R}^{n}right) ,mathcal{S}^{prime}left( mathbb{R}^{m}right) right) $ and $mathcal{S}^{prime}left( mathbb{R}^{n+m}right) $, we show that a large class of moderate linear mappings acting between the space $mathcal{G}_{mathcal{S}}left( mathbb{R}^{n}right) $ of Colombeau rapidly decreasing generalized functions and the space $mathcal{G}_{ au}left( mathbb{R}^{n}right) $ of temperate ones admits generalized integral representations, with kernels belonging to $mathcal{G}_{ au}left( mathbb{R}^{n+m}right) $. Furthermore, this result contains the classical one in the sense of the generalized distribution equality.
Math. Proc. Camb. Philos. Soc.

Droits : info:eu-repo/semantics/OpenAccess
hal-00019916
https://hal.archives-ouvertes.fr/hal-00019916
https://hal.archives-ouvertes.fr/hal-00019916/document
https://hal.archives-ouvertes.fr/hal-00019916/file/KernelSADelcroixfeb7th.pdf
ARXIV : math.FA/0603035

Éditeur(s) : HAL CCSD
Résumé : National audience
Designing systems able to interact with students in a natural manner is a complex and far from solved problem. A key aspect of natural interaction is the ability to understand and appropriately respond to human emotions. This paper details our response to the continuous Audio/Visual Emotion Challenge (AVEC'12) whose goal is to predict four affective signals describing human emotions. The proposed method uses Fourier spectra to extract multi-scale dynamic descriptions of signals characterizing face appearance, head movements and voice. We perform a kernel regression with very few representative samples selected via a supervised weighted-distance-based clustering, that leads to a high generalization power. We also propose a particularly fast regressor-level fusion framework to merge systems based on different modalities. Experiments have proven the efficiency of each key point of the proposed method and our results on challenge data were the highest among 10 international research teams.
IHM'14, 26e conférence francophone sur l'Interaction Homme-Machine
Lille, France

hal-01089628
https://hal.archives-ouvertes.fr/hal-01089628
https://hal.archives-ouvertes.fr/hal-01089628/document
https://hal.archives-ouvertes.fr/hal-01089628/file/p44-nicole.pdf