1,721,143 research outputs found
Method for producing solutions to a concrete multicriteria optimisation problem
No full textDemandeurs : Thales SA, Lehuédé Fabien, Grabisch Michel, Labreuche Christophe, Saveant Pierre ; Date de dépôt : 28/01/2003 ; N° de dépôt : FR20030000898The invention relates to a method of producing solutions to a concrete multicriteria optimisation problem. The inventive method consists in establishing several decision criteria and a preference relation based on said criteria between the solutions to the problem and modelling the problem to be resolved. The invention is characterised by the fact that: the solutions are obtained in a constructive manner using a tree search process; a tree search strategy has been established for each criterion; the strategies are alternated in order to find solutions of increasing quality; the strategies are selected dynamically according to the last solution found; the strategies are continuously alternated until a termination condition has been satisfied; and the last solution found before the verification of the termination condition is exhibited as the solution to the problem posed
Method for producing solutions to a concrete multicriteria optimisation problem
No full textDemandeurs : Thales SA, Lehuédé Fabien, Grabisch Michel, Labreuche Christophe, Saveant Pierre ; Date de dépôt : 28/01/2003 ; N° de dépôt : FR20030000898The invention relates to a method of producing solutions to a concrete multicriteria optimisation problem. The inventive method consists in establishing several decision criteria and a preference relation based on said criteria between the solutions to the problem and modelling the problem to be resolved. The invention is characterised by the fact that: the solutions are obtained in a constructive manner using a tree search process; a tree search strategy has been established for each criterion; the strategies are alternated in order to find solutions of increasing quality; the strategies are selected dynamically according to the last solution found; the strategies are continuously alternated until a termination condition has been satisfied; and the last solution found before the verification of the termination condition is exhibited as the solution to the problem posed
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Variations on the Author
Full text link“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
Appropriate Similarity Measures for Author Cocitation Analysis
Full text linkWe provide a number of new insights into the methodological discussion about author cocitation analysis. We first argue that the use of the Pearson correlation for measuring the similarity between authors’ cocitation profiles is not very satisfactory. We then discuss what kind of similarity measures may be used as an alternative to the Pearson correlation. We consider three similarity measures in particular. One is the well-known cosine. The other two similarity measures have not been used before in the bibliometric literature. Finally, we show by means of an example that our findings have a high practical relevance.information science;Pearson correlation;cosine;similarity measure;author cocitation analysis
Dispelling the Myths Behind First-author Citation Counts
Full text linkWe conducted a full-scale evaluative citation analysis study of scholars in the XML research field to explore just how different from each other author rankings resulting from different citation counting methods actually are, and to demonstrate the capability of emerging data and tools on the Web in supporting more realistic citation counting methods. Our results contest some common arguments for the continued
use of first-author citation counts in the evaluation of scholars, such as high correlations between author rankings by first-author citation counts and other citation
counting methods, and high costs of using more realistic citation counting methods that are not well-supported by the ISI databases. It is argued that increasingly available digital full text research papers make it possible for citation analysis studies to go beyond what the ISI databases have directly supported and to employ more
sophisticated methods
koamabayili/VECTRON-author-checklist: VECTRON author checklist
No full textWe have done our best to complete the author checklist relating to the use of animals in the hut study. Note that the objective for the hut study was to evaluate the IRS treatment applications for residual efficacy against Anopheles mosquitoes, including the local An. coluzzii mosquito population. Cows were only used to attract mosquitoes into the huts and no tests were carried out directly on the cows. The author checklist is intended for use with studies where experiments are carried out on animals, which is why we have had such difficulty in completing this for the hut study, as many of the questions do not relate to how the cows were used
Vers un apprentissage automatique fiable : exploiter les représentations multimodales, le goulot d’étranglement de l’information et la théorie des valeurs extrêmes (EVT)
No full textCette thèse de doctorat porte sur l'amélioration de la fiabilité de l'apprentissage automatique, en particulier pour les applications à forts enjeux. Les modèles d'apprentissage profond actuels, bien que très performants, restent difficiles à appréhender et à déployer de manière sûre en raison de leur opacité, de leur vulnérabilité aux attaques adverses, de leur sensibilité aux changements de distribution, et de leur inefficacité en contexte de données ou de ressources limitées. Pour surmonter ces limites, ce travail explore trois dimensions complémentaires : l'explicabilité, la robustesse et la frugalité.Sur le plan de l’explicabilité, il propose une méthode appelée CB2, qui introduit une forme d’interprétabilité par concepts pour les réseaux neuronaux profonds. CB2 s’appuie sur des embeddings multi-modaux et la théorie de la décision pour aligner les représentationsinternes du modèle sur des concepts compréhensibles par l’humain. Cette technique permet de mieux comprendre pourquoi un modèle produit une prédiction donnée et d’inspecter les biais potentiels dans le processus de décision. Par rapport aux méthodes post-hoc classiques, CB2 fournit des explications plus structurées et sémantiquement riches, validées sur plusieurs jeux de données en vision par ordinateur.Sur le plan de la robustesse, deux approches sont proposées pour renforcer la fiabilité des modèles. La première, nommée POSIB, est une méthode post-entraînement fondée sur le principe de l’Information Bottleneck, qui restructure l’espace latent du modèle afin de dissocier les caractéristiques informatives du bruit et des corrélations non pertinentes. Cela améliore la robustesse sans compromettre la précision prédictive. La seconde approche, appelée SPADE, traite de la détection des échantillons hors distribution (OOD) et des entrées adverses. SPADE exploite la théorie des valeurs extrêmes pour caractériser le comportement des queues de distribution latente, offrant ainsi une manière rigoureuse de détecter les entrées inconnues ou malveillantes et de s’abstenir de prédictions peu fiables. Les expériences menées sur différentes architectures et jeux de données montrent que SPADE atteint des performances de pointe en détection OOD et en défense contre les attaques adverses.Enfin, ce travail s’intéresse aussi à la frugalité, en reconnaissant que les modèles déployés dans des contextes industriels ou critiques fonctionnent souvent sous des contraintes sévères en données et en ressources de calcul. Pour relever ces défis, il développe des techniques d’apprentissage de représentations frugales qui optimisent le contenu informationnel des espaces latents, en ne conservant que les caractéristiques essentielles. Combiné avec le cadre de robustesse, il propose F-STUDENT, une version distillée des modèles robustes, qui compresse le réseau tout en préservant sa capacité à résister aux attaques adverses. Cette approche surpasse les méthodes classiques d’élagage en s’appuyant sur une distillation multi-étapes et une régularisation par goulot d’étranglement informationnel.Dans l’ensemble, cette thèse contribue à combler le fossé entre les avancées théoriques de l’apprentissage automatique moderne et les exigences pratiques du déploiement de l’IA dans des environnements critiques. En abordant conjointement l’explicabilité, la robustesse et la frugalité, elle propose un cadre complet pour le développement de systèmes d’apprentissage fiables, dignes de confiance dans des applications réelles à forts enjeux.This PhD research focuses on making machine learning more reliable, particularly for high-stakes applications. Current deep learning models, while highly performant, remain difficult to trust because of their opacity, vulnerability to adversarial attacks, sensitivity to distribution shifts, and inefficiency when faced with limited data or resources. To address these limitations, this work explores three complementary dimensions: explainability, robustness, and frugality. On the explainability side, it proposes a method called CB2, which introduces concept-based interpretability for deep neural networks. CB2 relies on multi-modal embeddings and decision theory to align internal model representations with human-interpretable concepts. This technique allows us to better understand why a model makes a given prediction and inspect potential biases in the decision process. Compared to traditional post-hoc attribution methods, CB2 offers more structured and semantically meaningful explanations, validated across several computer vision datasets. In terms of robustness, it proposes two approaches to enhance model reliability. The first is POSIB, a post-training method based on the Information Bottleneck principle, which reshapes a model's latent space to disentangle informative features from noise and irrelevant correlations. This improves robustness without sacrificing predictive accuracy. The second approach, called SPADE, addresses the detection of out-of-distribution samples and adversarial inputs. SPADE leverages Extreme Value Theory to characterize the tail behavior of latent distributions, providing a principled way to detect unfamiliar or malicious inputs and abstain from unreliable predictions. Experiments across different architectures and datasets show that SPADE achieves state-of-the-art performance in both OOD detection and adversarial defense. Finally, this work also focuses on frugality, recognizing that models deployed in industrial or safety-critical contexts often operate under severe data and computational constraints. To meet these challenges, it develops frugal representation learning techniques that optimize the information content of latent spaces, keeping only the most essential features. Combined with the robustness framework, it proposes F-STUDENT, a distilled version of robust models that compresses the network while preserving its ability to withstand adversarial attacks. This approach outperforms standard pruning methods by leveraging multi-step distillation and information bottleneck regularization. Overall, this PhD contributes to closing the gap between the theoretical progress of modern machine learning and the practical requirements of deploying AI in safety-critical environments. By addressing explainability, robustness, and frugality together, it offers a comprehensive framework for developing reliable machine learning systems that can be trusted in real-world high-stakes applications
- …
