1,720,971 research outputs found

    Learning higher-order functions for computation, memorization and control with artificial neural networks

    No full text
    Malgré des décennies de recherche, la compréhension des mécanismes d'apprentissage permettant aux réseaux de neurones biologiques de synthétiser l'expérience vécue pour acquérir de nouvelles compétences et d'articuler ces connaissances dans l’exécution de fonctions cognitives supérieures résiste encore a une description scientifique complète. Au même moment, les avancées du calcul informatique a grande échelle ont permis le développement de leur homologue artificiels, qui sous-tendent actuellement une révolution informationnelle. Ces modèles neuronaux simplifiés présentent toutefois d'importantes différences d'apprentissage en comparaison des réseaux naturels et apporte des explications insatisfaisantes sur les principes computationnels que l'évolution a favorisé au sein des ré- seaux biologiques. Le manque de compositionnalité et de systématicité constitue une limitation évidente du calcul neural artificiel et révèle son échec en tant que véritable paradigme de programmation fonctionnelle. Afin d'identifier les mécanismes susceptibles de permettre une intelligence machine plus abstractive, nous explorons dans cette thèse le problème de l'apprentissage de fonction d'ordre supérieur (i.e fonctions de fonctions) par réseaux de neurones artificiels. Nous proposons de marier les techniques d'apprentissage actuelles à ces réseaux vus comme fonctionnelles contrôlables, permettant de définir des modèles présentant des facultés calculatoires originales comme la mémorisation associative dynamique ou la régression d'opérateurs fonctionnels. En particulier, nous considérons le problème du méta-apprentissage, consistant à remplacer les méthodes manuelle d'ajustement fonctionnel par des méthodes d'adaptation elle-même apprises. En donnant aux réseaux de neurones la possibilité de construire leur propre stratégies d'apprentissage en fonction des données, de leur activité et de leur historique décisionnel, nous explorons de nouvelles formes de programme adaptatifs, trouvant des applications diverses en apprentissage "low-shot" ou en robotiques.After decades of research, the question of how biological neural networks synthesize experience to serve higher-level cognitive processes (concept acquisition, systematic decision making, evaluative thinking, creativity...) still resists to a complete scientific understanding. At the same time, advances in large scale computation have enabled the development of their artificial counterparts, which currently power a revolution in machine intelligence. These simplified neural models developed through a global scientific trial-and-error process, present however evident behavioral differences and limitations compared to biological neural networks and provide unsatisfactory to little insight into the computational principles that evolution condensed in biological neurons. One evident limitation is the lack of compositionality and systematicity in artificial neural learning, revealing its current failure as a true functional programming paradigm. In an attempt to identify key neural mechanisms able to support more abstractive machine intelligence, this thesis propose to explore the problem of learning higher-order functions (i.e functions of functions) with neural networks. Namely, we propose to marry modern machine learning techniques with the perspective of neural networks as controllable functionals, unearthing original computational mechanisms able to support powerful faculties such as dynamic associative memorization or functional operator regression. In particular, building on the clear epistemological trend that hand-designed methods are eventually replaced by computerized and self-executing solutions, we will consider the specific higher-order problem of learning to learn, i.e meta-learning. We show that granting neural networks the aptitude to design learning strategies with minimal solution constraints as a function of data, activity or choice history, can help explore new forms of adaptive programs, with application to low-shot learning or robotic control

    Going Beyond Counting First Authors in Author Co-citation Analysis

    Get PDF
    The 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

    Apprentissage de fonctions d'ordre supérieur pour le calcul, la mémorisation et le contrôle par réseaux de neurones artificiels

    No full text
    After decades of research, the question of how biological neural networks synthesize experience to serve higher-level cognitive processes (concept acquisition, systematic decision making, evaluative thinking, creativity...) still resists to a complete scientific understanding. At the same time, advances in large scale computation have enabled the development of their artificial counterparts, which currently power a revolution in machine intelligence. These simplified neural models developed through a global scientific trial-and-error process, present however evident behavioral differences and limitations compared to biological neural networks and provide unsatisfactory to little insight into the computational principles that evolution condensed in biological neurons. One evident limitation is the lack of compositionality and systematicity in artificial neural learning, revealing its current failure as a true functional programming paradigm. In an attempt to identify key neural mechanisms able to support more abstractive machine intelligence, this thesis propose to explore the problem of learning higher-order functions (i.e functions of functions) with neural networks. Namely, we propose to marry modern machine learning techniques with the perspective of neural networks as controllable functionals, unearthing original computational mechanisms able to support powerful faculties such as dynamic associative memorization or functional operator regression. In particular, building on the clear epistemological trend that hand-designed methods are eventually replaced by computerized and self-executing solutions, we will consider the specific higher-order problem of learning to learn, i.e meta-learning. We show that granting neural networks the aptitude to design learning strategies with minimal solution constraints as a function of data, activity or choice history, can help explore new forms of adaptive programs, with application to low-shot learning or robotic control.Malgré des décennies de recherche, la compréhension des mécanismes d'apprentissage permettant aux réseaux de neurones biologiques de synthétiser l'expérience vécue pour acquérir de nouvelles compétences et d'articuler ces connaissances dans l’exécution de fonctions cognitives supérieures résiste encore a une description scientifique complète. Au même moment, les avancées du calcul informatique a grande échelle ont permis le développement de leur homologue artificiels, qui sous-tendent actuellement une révolution informationnelle. Ces modèles neuronaux simplifiés présentent toutefois d'importantes différences d'apprentissage en comparaison des réseaux naturels et apporte des explications insatisfaisantes sur les principes computationnels que l'évolution a favorisé au sein des ré- seaux biologiques. Le manque de compositionnalité et de systématicité constitue une limitation évidente du calcul neural artificiel et révèle son échec en tant que véritable paradigme de programmation fonctionnelle. Afin d'identifier les mécanismes susceptibles de permettre une intelligence machine plus abstractive, nous explorons dans cette thèse le problème de l'apprentissage de fonction d'ordre supérieur (i.e fonctions de fonctions) par réseaux de neurones artificiels. Nous proposons de marier les techniques d'apprentissage actuelles à ces réseaux vus comme fonctionnelles contrôlables, permettant de définir des modèles présentant des facultés calculatoires originales comme la mémorisation associative dynamique ou la régression d'opérateurs fonctionnels. En particulier, nous considérons le problème du méta-apprentissage, consistant à remplacer les méthodes manuelle d'ajustement fonctionnel par des méthodes d'adaptation elle-même apprises. En donnant aux réseaux de neurones la possibilité de construire leur propre stratégies d'apprentissage en fonction des données, de leur activité et de leur historique décisionnel, nous explorons de nouvelles formes de programme adaptatifs, trouvant des applications diverses en apprentissage "low-shot" ou en robotiques

    Apprentissage de fonctions d'ordre supérieur pour le calcul, la mémorisation et le contrôle par réseaux de neurones artificiels

    No full text
    After decades of research, the question of how biological neural networks synthesize experience to serve higher-level cognitive processes (concept acquisition, systematic decision making, evaluative thinking, creativity...) still resists to a complete scientific understanding. At the same time, advances in large scale computation have enabled the development of their artificial counterparts, which currently power a revolution in machine intelligence. These simplified neural models developed through a global scientific trial-and-error process, present however evident behavioral differences and limitations compared to biological neural networks and provide unsatisfactory to little insight into the computational principles that evolution condensed in biological neurons. One evident limitation is the lack of compositionality and systematicity in artificial neural learning, revealing its current failure as a true functional programming paradigm. In an attempt to identify key neural mechanisms able to support more abstractive machine intelligence, this thesis propose to explore the problem of learning higher-order functions (i.e functions of functions) with neural networks. Namely, we propose to marry modern machine learning techniques with the perspective of neural networks as controllable functionals, unearthing original computational mechanisms able to support powerful faculties such as dynamic associative memorization or functional operator regression. In particular, building on the clear epistemological trend that hand-designed methods are eventually replaced by computerized and self-executing solutions, we will consider the specific higher-order problem of learning to learn, i.e meta-learning. We show that granting neural networks the aptitude to design learning strategies with minimal solution constraints as a function of data, activity or choice history, can help explore new forms of adaptive programs, with application to low-shot learning or robotic control.Malgré des décennies de recherche, la compréhension des mécanismes d'apprentissage permettant aux réseaux de neurones biologiques de synthétiser l'expérience vécue pour acquérir de nouvelles compétences et d'articuler ces connaissances dans l’exécution de fonctions cognitives supérieures résiste encore a une description scientifique complète. Au même moment, les avancées du calcul informatique a grande échelle ont permis le développement de leur homologue artificiels, qui sous-tendent actuellement une révolution informationnelle. Ces modèles neuronaux simplifiés présentent toutefois d'importantes différences d'apprentissage en comparaison des réseaux naturels et apporte des explications insatisfaisantes sur les principes computationnels que l'évolution a favorisé au sein des ré- seaux biologiques. Le manque de compositionnalité et de systématicité constitue une limitation évidente du calcul neural artificiel et révèle son échec en tant que véritable paradigme de programmation fonctionnelle. Afin d'identifier les mécanismes susceptibles de permettre une intelligence machine plus abstractive, nous explorons dans cette thèse le problème de l'apprentissage de fonction d'ordre supérieur (i.e fonctions de fonctions) par réseaux de neurones artificiels. Nous proposons de marier les techniques d'apprentissage actuelles à ces réseaux vus comme fonctionnelles contrôlables, permettant de définir des modèles présentant des facultés calculatoires originales comme la mémorisation associative dynamique ou la régression d'opérateurs fonctionnels. En particulier, nous considérons le problème du méta-apprentissage, consistant à remplacer les méthodes manuelle d'ajustement fonctionnel par des méthodes d'adaptation elle-même apprises. En donnant aux réseaux de neurones la possibilité de construire leur propre stratégies d'apprentissage en fonction des données, de leur activité et de leur historique décisionnel, nous explorons de nouvelles formes de programme adaptatifs, trouvant des applications diverses en apprentissage "low-shot" ou en robotiques

    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 PDF
    We 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 PDF
    We 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

    Author Index

    No full text
    Nao informado

    koamabayili/VECTRON-author-checklist: VECTRON author checklist

    No full text
    We 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
    corecore