1,720,957 research outputs found
Analyse en Composantes Indépendantes Multidimensionnelles via des cumulants d’ordres variés
Get PDFThe author deals with the problem of multidimensional independent component analysis (MICA) which is the natural generalization of the ordinary problem of independent component analysis (ICA). First, in order to facilitate the use of higher-order cumulants, we present new formulas for the cumulant matrices of a real random vector from its moment matrices. In addition to the usual matrix operations, these formulas use only the Kronecker product, the vec operator and some commutation matrices. These formulas lend themselves to examine more closely the specific structures of cumulant matrices and provide results on the ranks of these matrices that characterize the dependence between random variables composing the random vector. The main practical interest of our matrix formulas lies in much easier cumulant evaluation and faster computation than the conventional method based on repeated use of the Leonov and Shiryaev formulas. In the second part of this thesis, we show that under the usual assumptions of the independent multidimensional component analysis, contracted cumulant matrices at any statistical order are all block diagonalizable in the same basis. We derive an algorithm for solving MICA by block diagonalization and compare the results obtained to the orders 3-6, between them and with other methods, on several synthetic signals. Simple examples are developed to justify the need to combine different levels to ensure the best separation. We also prove that the easiest case to deal with is the case of mixtures of sources that have different dimensions. In the last part of this thesis we propose a set of methods that operate only the higher- order statistics. Under certain additional assumptions, these methods are shown to completely solve the noisy MICA problem without second-order whitening by joint block diagonalization of a cumulant matrices set coming from statistics of orders strictly higher than four. A comparison with the second-order based whitening MICA methods for the separation of fetal and maternal electrical activities (measured using three electrodes placed on the mother’s abdomen) shows that this new family is better suited to this application : it allow an almost perfect separation of these two contributions.L’auteur s’intéresse au problème de l’analyse en composantes indépendantes multidimensionnelles (ACIM) qui est la généralisation naturelle du problème ordinaire de l’analyse en composantes indépendantes (ACI). Tout d’abord, afin de faciliter l’utilisation des cumulants des ordres supérieurs, nous présentons de nou- velles formules pour le calcul matriciel des matrices de cumulants d’un vecteur aléatoire réel à partir de ses matrices de moments. Outre les opérations matricielles usuelles, ces formules utilisent uniquement le produit de Kronecker, l’opérateur vec et des matrices de commutation. Nous pouvons immédiatement à partir de ces formules examiner de plus près les structures particulières des matrices de cumulants et ainsi donner des résultats sur les rangs de ces matrices qui caractérisent la dépendance entre les variables aléatoires constituant le vecteur aléatoire. L’intérêt pratique principal de nos formules matricielles réside certainement dans une évaluation des cumulants beaucoup plus aisée et rapide qu’avec la méthode usuelle basée sur une utilisation répétée des formules de Leonov et Shiryaev. Dans la deuxième partie de cette thèse, nous montrons que sous les hypothèses usuelles de l’analyse en composantes indépendantes mul- tidimensionnelles, les matrices de cumulants contractées à un ordre statistique quelconque sont toutes bloc-diagonalisables dans la même base. Nous en déduisons des algorithmes de résolution d’ACIM par bloc-diagonalisation conjointe et comparons les résultats obtenus aux ordres 3 à 6, entre eux et avec d’autres méthodes, sur quelques signaux synthétiques. Des exemples simples ont élaborés afin de justifier la nécessité de combiner des ordres différents pour garantir la meilleure séparation. Nous prouvons aussi que le cas le plus simple à traiter est celui de mélanges de sources qui ont différentes dimensions. Dans la dernière partie de cette thèse nous proposons une famille de méthodes qui exploitent uniquement les sta- tistiques d’ordres supérieurs à deux. Sous certaines hypothèses supplémentaires, ces méthodes permettent après un blanchiment d’ordre quatre des observations de résoudre complètement le problème ACIM bruité en bloc diagonalisant conjointement un ensemble de matrices de cumulants issues des statistiques d’ordres supérieurs strictement à quatre. Une comparaison avec les méthodes ACIM à blanchiment d’ordre deux pour la séparation des activités électriques foetale et maternelle (mesurées via trois électrodes placées sur l’abdomen de la mère) montre que cette nouvelle famille est mieux adaptée à cette application : elles permettent une séparation quasi parfaite de ces deux contributions
Analyse en Composantes Indépendantes Multidimensionnelles via des cumulants d’ordres variés
No full textThe author deals with the problem of multidimensional independent component analysis (MICA) which is the natural generalization of the ordinary problem of independent component analysis (ICA). First, in order to facilitate the use of higher-order cumulants, we present new formulas for the cumulant matrices of a real random vector from its moment matrices. In addition to the usual matrix operations, these formulas use only the Kronecker product, the vec operator and some commutation matrices. These formulas lend themselves to examine more closely the specific structures of cumulant matrices and provide results on the ranks of these matrices that characterize the dependence between random variables composing the random vector. The main practical interest of our matrix formulas lies in much easier cumulant evaluation and faster computation than the conventional method based on repeated use of the Leonov and Shiryaev formulas. In the second part of this thesis, we show that under the usual assumptions of the independent multidimensional component analysis, contracted cumulant matrices at any statistical order are all block diagonalizable in the same basis. We derive an algorithm for solving MICA by block diagonalization and compare the results obtained to the orders 3-6, between them and with other methods, on several synthetic signals. Simple examples are developed to justify the need to combine different levels to ensure the best separation. We also prove that the easiest case to deal with is the case of mixtures of sources that have different dimensions. In the last part of this thesis we propose a set of methods that operate only the higher- order statistics. Under certain additional assumptions, these methods are shown to completely solve the noisy MICA problem without second-order whitening by joint block diagonalization of a cumulant matrices set coming from statistics of orders strictly higher than four. A comparison with the second-order based whitening MICA methods for the separation of fetal and maternal electrical activities (measured using three electrodes placed on the mother’s abdomen) shows that this new family is better suited to this application : it allow an almost perfect separation of these two contributions.L’auteur s’intéresse au problème de l’analyse en composantes indépendantes multidimensionnelles (ACIM) qui est la généralisation naturelle du problème ordinaire de l’analyse en composantes indépendantes (ACI). Tout d’abord, afin de faciliter l’utilisation des cumulants des ordres supérieurs, nous présentons de nou- velles formules pour le calcul matriciel des matrices de cumulants d’un vecteur aléatoire réel à partir de ses matrices de moments. Outre les opérations matricielles usuelles, ces formules utilisent uniquement le produit de Kronecker, l’opérateur vec et des matrices de commutation. Nous pouvons immédiatement à partir de ces formules examiner de plus près les structures particulières des matrices de cumulants et ainsi donner des résultats sur les rangs de ces matrices qui caractérisent la dépendance entre les variables aléatoires constituant le vecteur aléatoire. L’intérêt pratique principal de nos formules matricielles réside certainement dans une évaluation des cumulants beaucoup plus aisée et rapide qu’avec la méthode usuelle basée sur une utilisation répétée des formules de Leonov et Shiryaev. Dans la deuxième partie de cette thèse, nous montrons que sous les hypothèses usuelles de l’analyse en composantes indépendantes mul- tidimensionnelles, les matrices de cumulants contractées à un ordre statistique quelconque sont toutes bloc-diagonalisables dans la même base. Nous en déduisons des algorithmes de résolution d’ACIM par bloc-diagonalisation conjointe et comparons les résultats obtenus aux ordres 3 à 6, entre eux et avec d’autres méthodes, sur quelques signaux synthétiques. Des exemples simples ont élaborés afin de justifier la nécessité de combiner des ordres différents pour garantir la meilleure séparation. Nous prouvons aussi que le cas le plus simple à traiter est celui de mélanges de sources qui ont différentes dimensions. Dans la dernière partie de cette thèse nous proposons une famille de méthodes qui exploitent uniquement les sta- tistiques d’ordres supérieurs à deux. Sous certaines hypothèses supplémentaires, ces méthodes permettent après un blanchiment d’ordre quatre des observations de résoudre complètement le problème ACIM bruité en bloc diagonalisant conjointement un ensemble de matrices de cumulants issues des statistiques d’ordres supérieurs strictement à quatre. Une comparaison avec les méthodes ACIM à blanchiment d’ordre deux pour la séparation des activités électriques foetale et maternelle (mesurées via trois électrodes placées sur l’abdomen de la mère) montre que cette nouvelle famille est mieux adaptée à cette application : elles permettent une séparation quasi parfaite de ces deux contributions
Concise formulae for the cumulant matrices of a random vector
No full text18,Concise formulae are given for the cumulant matrices of a random vector up to order 6. In addition to usual matrix operations, they involve only the Kronecker product, the vec operator, and the commutation matrix. Orders 5 and 6 are provided here for the first time; the same method as provided in the paper can be applied to compute higher orders. An immediate consequence of these formulae is to return 1) upper bounds on the rank of the cumulant matrices and 2) the expression of the sixth-order moment matrix of a Gaussian vector. Due to their conciseness, the proposed formulae also have a computational advantage as compared to the repeated use of Leonov and Shiryaev formula
Multidimensional Independent Component Analysis with Higher-order cumulant matrices for vector sources with possibly differing dimensions
No full text29The paper addresses the separation of multidimensional sources, with possibly different dimensions, by means of higher-order cumulant matrices. First, it is rigorously proved, in a general setting, that contracted cumulant matrices of any order are all block-diagonalizable in the same basis. Second, a family of joint block-diagonalization algorithms is proposed that separate multidimensional sources by combining contracted cumulant matrices of arbitrary orders. Third, a specific solution is given to determine the source dimensions when they are unknown but all different. The performances of the proposed algorithms are compared between them and with algorithms of the literature based on orders 3 and 6
Concise formulae for the cumulant matrices of a random vector
No full textInternational audienceConcise formulae are given for the cumulant matrices of a real-valued (zero-mean) random vector up to order 6. In addition to usual matrix operations, they involve only the Kronecker product, the vec operator, and the commutation matrix. Orders 5 and 6 are provided here for the first time; the same method as provided in the paper can be applied to compute higher orders. An immediate consequence of these formulae is to return 1) upper bounds on the rank of the cumulant matrices and 2) the expression of the sixth-order moment matrix of a Gaussian vector. Due to their conciseness, the proposed formulae also have a computational advantage as compared to the repeated use of Leonov and Shiryaev formula
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
A robust forecasting framework based on the Kalman filtering approach with a twofold parameter tuning procedure: Application to solar and photovoltaic prediction
Get PDFInternational audienceThis paper presents a framework which relies on the linear dynamical Kalman filter to perform a reliable prediction for solar and photovoltaic production. The method is convenient for real-time forecasting and we describe its use to perform these predictions for different time horizons, between one minute and one hour ahead. The dataset used is a set of measurements of solar irradiance and PV power production measured in a subtropical zone: Guadeloupe. In this zone, fluctuating meteorological conditions can occur, with highly variable atmospheric events having severe impact in the solar irradiance and the PV power. In such conditions, heterogeneous ramp events are observed making difficult to control and manage these sources of energy. The present work hopes to build a suitable statistical method, based on bayesian inference and state-space modeling, able to predict the evolution of solar radiation and PV production. We develop a forecast method based on the Kalman filter combined with a robust parameter estimation procedure built with an Auto Regressive model or with an Expectation–Maximisation algorithm. The model is built to run with univariate or multivariate data according to their availability. The model is used here to forecast the univariate solar and PV data and also PV with exogenous data such as cloud cover and air temperature. The accuracy of this technique is studied with a set of performance criterion including the root mean square error and the mean bias error. We compare the results for the different tests performed, from one minute to one hour ahead, to the simple persistence model. The performance of our technique exceeds by far the traditional persistence model with a skill score improvement around 39% and 31%, respectively for PV production and GHI, for one hour ahead forecast
- …
