Éditeur(s) : Paris : Librairie militaire L. Fournier

Droits : Domaine public
Provenance : Université de Nantes. Service commun de la documentation
Permalien : http://www.manioc.org/patrimon/NAN13041
NAN13041

Éditeur(s) : [S.l.] : Delagrave
Siècle(s) traité(s) : 19
Droits : Domaine public
Provenance : Université des Antilles. Service commun de la documentation
Permalien : http://www.manioc.org/patrimon/HASH010d08988893e8073d1ea073
HASH010d08988893e8073d1ea073

Éditeur(s) : HAL CCSD springer Link
Résumé : International audience
Experimental Astronomy

hal-01094521
https://hal.archives-ouvertes.fr/hal-01094521
BIBCODE : 2014ExA...tmp...41R
DOI : 10.1007/s10686-014-9383-4
ARXIV : 1310.0696

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

Résumé : Publiée dans le cadre du premier Congrès international d'histoire coloniale à Paris
Siècle(s) traité(s) : 20
Droits : Domaine public
Provenance : Ville de Pointe-à-Pitre. Réseau des bibliothèques
Permalien : http://www.manioc.org/patrimon/PAP11103
PAP11103

Éditeur(s) : HAL CCSD Geological Society of America
Résumé : International audience
Fluid-rock interactions can control earthquake nucleation and the evolution of earthquake sequences. Experimental studies of fault frictional properties in the presence of fluid can provide unique insights into these interactions. We report the first results from experiments performed on cohesive silicate-bearing rocks (microgabbro) in the presence of pressurized pore fluids (H2O, drained conditions) at realistic seismic deformation conditions. The experimental data are compared with those recently obtained from carbonate-bearing rocks (Carrara marble). Contrary to theoretical arguments, and consistent with the interpretation of some field observations, we show that frictional melting of a microgabbro develops in the presence of water. In microgabbro, the initial weakening mechanism (flash melting of the asperities) is delayed in the presence of water; conversely, in calcite marble the weakening mechanism (brittle failure of the asperities) is favored. This opposite behavior highlights the importance of host-rock composition in controlling dynamic (frictional) weakening in the presence of water: cohesive carbonate-bearing rocks are more prone to slip in the presence of water, whereas the presence of water might delay or inhibit the rupture nucleation and propagation in cohesive silicate-bearing rocks.
ISSN: 0091-7613

hal-01054327
https://hal.archives-ouvertes.fr/hal-01054327
DOI : 10.1130/G34916.1

Éditeur(s) : HAL CCSD American Geophysical Union
Résumé : International audience
We performed 31 friction experiments on glassy basalts (GB) and glass-free basalts (GFB) at slip rates up to 6.5 m s−1 and normal stress up to 40 MPa (seismic conditions). Frictional weakening was associated to bulk frictional melting and lubrication. The weakening distance (Dw) was about 3 times shorter in GB than in GFB, but the steady state friction was systematically higher in GB than in GFB. The shorter Dw in GB may be explained by the thermal softening occurring at the glass transition temperature (Tg ~500°C), which is lower than the bulk melting temperature (Tm ~1250°C) of GFB. Postexperiment microanalyses suggest that the larger crystal fraction measured in GB melts results in the higher steady state friction value compared to the GFB melts. The effect of interstitial glass is to facilitate frictional instability and rupture propagation in GB with respect to GFB.
ISSN: 0094-8276

hal-01054139
https://hal.archives-ouvertes.fr/hal-01054139
DOI : 10.1002/2013GL058601

Éditeur(s) : HAL CCSD Institute of Electrical and Electronics Engineers
Résumé : International audience
Tailoring nearest neighbors algorithms to boosting is an important problem. Recent papers study an approach, UNN, which provably minimizes particular convex surrogates under weak assumptions. However, numerical issues make it necessary to experimentally tweak parts of the UNN algorithm, at the possible expense of the algorithm's convergence and performance. In this paper, we propose a lightweight alternative algorithm optimizing proper scoring rules from a very broad set, and establish formal convergence rates under the boosting framework that surprisingly compete with those known for UNN. It is an adaptive Newton-Raphson algorithm, which belongs to the same lineage as the popular Gentle Adaboost. To the best of our knowledge, no such boosting-compliant convergence rates were previously known for these algorithms. We provide experiments on a dozen domains, including the challenging Caltech and SUN computer vision databases. They display that GNNB significantly outperforms UNN, both in terms of convergence rate and quality of the solution obtained, and GNNB provides a simple and efficient contender to techniques that can be used on very large domains, like stochastic gradient descent -- for which little is known to date. Experiments include a divide-and-conquer improvement of GNNB which exploits the link with proper scoring rules optimization.
ISSN: 0162-8828

hal-00958809
https://hal.archives-ouvertes.fr/hal-00958809
https://hal.archives-ouvertes.fr/hal-00958809/document
https://hal.archives-ouvertes.fr/hal-00958809/file/PAMI.pdf
DOI : 10.1109/TPAMI.2014.2307877

Éditeur(s) : HAL CCSD Springer
Résumé : Third International Workshop, MLMI 2012, Held in Conjunction with MICCAI 2012
International audience
Universal Nearest Neighbours (unn) is a classifier recently proposed, which can also effectively estimates the posterior probability of each classification act. This algorithm, intrinsically binary, requires the use of a decomposition method to cope with multiclass problems, thus reducing their complexity in less complex binary subtasks. Then, a reconstruction rule provides the final classification. In this paper we show that the application of unn algorithm in conjunction with a reconstruction rule based on the posterior probabilities provides a classification scheme robust among different biomedical image datasets. To this aim, we compare unn performance with those achieved by Support Vector Machine with two different kernels and by a k Nearest Neighbours classifier, and applying two different reconstruction rules for each of the aforementioned classification paradigms. The results on one private and five public biomedical datasets show satisfactory performance.
Machine Learning in Medical Imaging
Nice, France

hal-00958909
https://hal.archives-ouvertes.fr/hal-00958909
https://hal.archives-ouvertes.fr/hal-00958909/document
https://hal.archives-ouvertes.fr/hal-00958909/file/Miccai.pdf
DOI : 10.1007/978-3-642-35428-1_15

Éditeur(s) : HAL CCSD Cambridge University Press (CUP)
Résumé : International audience
Background. Adverse psychosocial working environments characterized by job strain (the combination of high demands and low control at work) are associated with an increased risk of depressive symptoms among employees, but evidence on clinically diagnosed depression is scarce. We examined job strain as a risk factor for clinical depression. Method. We identified published cohort studies from a systematic literature search in PubMed and PsycNET and obtained 14 cohort studies with unpublished individual-level data from the Individual-Participant-Data Meta-analysis in Working Populations (IPD-Work) Consortium. Summary estimates of the association were obtained using random-effects models. Individual-level data analyses were based on a pre-published study protocol. Results. We included six published studies with a total of 27 461 individuals and 914 incident cases of clinical depression. From unpublished datasets we included 120 221 individuals and 982 first episodes of hospital-treated clinical depression. Job strain was associated with an increased risk of clinical depression in both published [relative risk (RR) = 1.77, 95% confidence interval (CI) 1.47-2.13] and unpublished datasets (RR = 1.27, 95% CI 1.04-1.55). Further individual participant analyses showed a similar association across sociodemographic subgroups and after excluding individuals with baseline somatic disease. The association was unchanged when excluding individuals with baseline depressive symptoms (RR = 1.25, 95% CI 0.94-1.65), but attenuated on adjustment for a continuous depressive symptoms score (RR = 1.03, 95% CI 0.81-1.32). Conclusions. Job strain may precipitate clinical depression among employees. Future intervention studies should test whether job strain is a modifiable risk factor for depression.
ISSN: 0033-2917

hal-01560152
https://hal-univ-rennes1.archives-ouvertes.fr/hal-01560152
DOI : 10.1017/S003329171600355X
PUBMED : 28122650

Éditeur(s) : Kjobenhavn : H. Hagerups Boghandel
Siècle(s) traité(s) : 19
Droits : Domaine public
Provenance : Université Toulouse III - Paul Sabatier. Service Commun de la Documentation
Permalien : http://www.manioc.org/patrimon/TLS17095
TLS17095

Éditeur(s) : Londres : P. S. King and Son
Siècle(s) traité(s) : 19
Provenance : Université Caen Normandie. Service commun de la documentation
Permalien : http://www.manioc.org/patrimon/CAE19037
CAE19037