1,720,971 research outputs found
Transport optimal structurel pour l'adaptation au domaine avec garanties théoriques
No full textLa théorie du transport optimal permet non seulement de définir une distance entre les mesures de probabilité, mais offre également un moyen géométrique de transporter un ensemble de points vers un autre selon le principe du moindre effort. Ce double aspect a laissé la porte grande ouverte pour les applications en adaptation de domaine, une branche de l'apprentissage statistique qui tient compte du changement de distributions entre les données d'apprentissage et les données de test, respectivement appelées domaines source et cible. Toutefois, il existe souvent dans les deux domaines un biais structurel sur la représentation des données ou des structures latentes qui ne sont pas prises en compte par la formulation classique du transport optimal, et l'incapacité à incorporer pleinement ces structures peut entraver le succès de l'adaptation de domaine. Cette thèse présente plusieurs approches pour incorporer les informations structurelles au sein du transport optimal. La première contribution s'appuie sur une formulation hiérarchique du transport optimal pour aligner les structures sources et cibles. Les structures sources sont formées instinctivement en regroupant les données en classes selon leurs étiquettes, tandis que l'apprentissage des structures cachées dans le domaine cible est réduit au problème d'apprentissage de mesures de probabilité via le barycentre de Wasserstein, dont nous prouvons l'équivalence avec le clustering spectral. Notre deuxième contribution est une analyse théorique de l'adaptation de domaine à travers le transport optimal hiérarchique, où nous fournissons des bornes de généralisation pour trois scénarios, à savoir, l'adaptation de domaine non supervisé, semi-supervisé et multi-sources. Ces bornes de généralisation sont basées sur une nouvelle mesure de divergence que nous appelons la distance de Wasserstein Hiérarchique, qui indique, sous des hypothèses modérées, quelles structures doivent être alignées pour mener à une adaptation réussie. Dans notre troisième contribution, nous élargissons le cadre d'apprentissage des structures cibles en dehors du clustering, en développant une approche de propagation de labels basée sur le transport optimal. L'intérêt du transport optimal dans ce contexte est de capturer la géométrie de l'espace d'entrée dans son intégralité. Cette approche effectue une propagation incrémentale de labels, contrôlée par un score qui surveille la certitude des prédictions. Enfin, en s'appuyant sur ce nouvel algorithme de propagation de labels, nous présentons la dernière contribution, qui permet de créer de manière progressive des structures sources augmentées, permettant l'apprentissage d'une suite de sous-espaces latents domaine-invariants et discriminants, au sein desquels il devient facile d'étiqueter graduellement les données du domaine cible.Optimal transport theory not only defines a distance between probability measures but also provides a geometric way to transport a set of points to another according to the principle of least effort. This dual aspect has left the door wide open for applications in domain adaptation, a subfield of statistical learning theory that takes into account the change in distributions between training and test data, respectively called source and target domains. However, there is often a structural bias on the data representation or latent structures in both domains that are not captured by the classical optimal transport formulation, and the inability to fully incorporate these structures can hinder the success of domain adaptation. This thesis presents several approaches to incorporating structural information into the optimal transport problem. The first contribution relies on a hierarchical formulation of optimal transport to align source and target structures. The source structures are formed instinctively by grouping data into classes according to their labels while learning hidden structures in the target domain is reduced to the problem of learning probability measures through Wasserstein barycenter, which we prove to be equivalent to spectral clustering. Our second contribution is a new theoretical framework of domain adaptation through hierarchical optimal transport, where we provide generalization bounds for three scenarios, namely, unsupervised, semi-supervised, and multi-source domain adaptation. These generalization bounds are based on a new divergence measure that we call Hierarchical Wasserstein distance, indicating, under mild assumptions, which structures need to be aligned to lead to successful adaptation. In our third contribution, we extend the framework of learning target structures outside of clustering, by developing a label propagation approach based on optimal transport. The appeal of optimal transport in this context is to capture the geometry of the input space in its entirety. This approach performs incremental label propagation, controlled by a score that watches over the certainty of predictions. Finally, based on this new label propagation algorithm, we present the last contribution, which allows the progressive creation of augmented source structures, allowing to learn a sequence of latent domain-invariant and discriminative subspaces, within which it becomes easy to gradually label the target data
Clustering dans les réseaux attribués : Application aux systèmes de recommandation
Get PDFIn complex networks analysis field, much effort has been focused on identifying graphs communities of related nodes with dense internal connections and few external connections. In addition to node connectivity information that are mostly composed by different types of links, most real-world networks contains also node and/or edge associated attributes which can be very relevant during the learning process to find out the groups of nodes i.e. communities. In this case, two types of information are available : graph data to represent the relationship between objects and attributes information to characterize the objects i.e nodes. Classic community detection and data clustering techniques handle either one of the two types but not both. Consequently, the resultant clustering may not only miss important information but also lead to inaccurate findings. Therefore, various methods have been developed to uncover communities in networks by combining structural and attribute information such that nodes in a community are not only densely connected, but also share similar attribute values. Such graph-shape data is often referred to as attributed graph.This thesis focuses on developing algorithms and models for attributed graphs. Specifically, I focus in the first part on the different types of edges which represent different types of relations between vertices. I proposed a new clustering algorithms and I also present a redefinition of principal metrics that deals with this type of networks.Then, I tackle the problem of clustering using the node attribute information by describing a new original community detection algorithm that uncover communities in node attributed networks which use structural and attribute information simultaneously. At last, I proposed a collaborative filtering model in which I applied the proposed clustering algorithms.Au cours de la dernière décennie, les réseaux (les graphes) se sont révélés être un outil efficace pour modéliser des systèmes complexes. La problématique de détection de communautés est une tâche centrale dans l’analyse des réseaux complexes. La majeur partie des travaux dans ce domaine s’intéresse à la structure topologique des réseaux. Cependant, dans plusieurs cas réels, les réseaux complexes ont un ensemble d’attributs associés aux nœuds et/ou aux liens. Ces réseaux sont dites : réseaux attribués. Mes activités de recherche sont basées principalement sur la détection des communautés dans les réseaux attribués. Pour aborder ce problème, on s’est intéressé dans un premier temps aux attributs relatifs aux liens, qui sont un cas particulier des réseaux multiplexes. Un multiplex est un modèle de graphe multi-relationnel. Il est souvent représenté par un graphe multi-couches. Chaque couche contient le même ensemble de nœuds mais encode une relation différente. Dans mes travaux de recherche, nous proposons une étude comparative des différentes approches de détection de communautés dans les réseaux multiplexes. Cette étude est faite sur des réseaux réels. Nous proposons une nouvelle approche centrée "graine" pour la détection de communautés dans les graphes multiplexes qui a nécessité la redéfinition des métriques de bases des réseaux complexes au cas multiplex. Puis, nous proposons une approche de clustering dans les réseaux attribués qui prend en considération à la fois les attributs sur les nœuds et sur les liens. La validation de mes approches a été faite avec des indices internes et externes, mais aussi par une validation guidée par un système de recommandation que nous avons proposé et dont la détection de communautés est sa tâche principale. Les résultats obtenus sur ces approches permettent d’améliorer la qualité des communautés détectées en prenant en compte les informations sur les attributs du réseaux. De plus, nous offrons des outils d’analyse des réseaux attribués sous le langage de programmation R
Going Beyond Counting First Authors in Author Co-citation Analysis
Get PDFThe 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
Get PDF“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
Get PDFWe 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
Get PDFWe 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
Author-wise bibliometric analysis based on entropy.
No full textAuthor-wise bibliometric analysis based on entropy.</p
- …
