Dieser Bericht präsentiert RACR-NET, eine Schnittstelle der Referenzattributgrammatik-gesteuerten Graphersetzungsbibliothek RACR für C#. RACR-NET ermöglicht die Nutzung der deklarativen, dynamischen Sprachspezifikations-, Instanziierungs- und Auswertungsmeachanismen der RACR Scheme-Bibliothek in der objektorientierten Programmierung. Dies umfasst insbesondere die automatische inkrementelle Auswertung attributbasierter semantischer Analysen und somit das automatische Cachen parametrisierter Funktionsmethoden. Graphersetzungen entsprechen hierbei Zustandsänderungen von Objektinstanzen und der Invalidierung abgeleiteter Berechnungen. Schwerpunkt dieses Berichts ist die objektorientierte Programmierschnittstelle von RACR-NET, dessen praktische Anwendung und Implementierung. Der Bericht ist ein Referenzhandbuch für RACR-NET Anwender und Entwickler.

The aim of this thesis was to summarize current findings about exercise testing. The main goal was to evaluate protocols used in daily practice, their indications and benefits, comparability and parameters that can influence the results. The purpose of the practical part was to compare the maximal values of physiological parameters (heart rate, VO2, minute ventilation, RER, power output and rated perceived exertion) for the two most commonly used protocols for bicycle ergometry tests in the Czech Republic. We compared ramp (1/3W.kg-1 .min-1 ) and step (1W.kg-1 every 3 min) incremental protocols to exhaustion. Thirteen young, healthy and physically fit subjects (9 males and 4 women aged from 20 to 31 years) underwent two tests in separate occasions, during which they performed the protocols in random order. Paired t-test was used to statisticaly analyze the data. We found no statisticaly significant differences in these maximal values except for the exercise test time. We deduced some practical benefits of each protokol from obtained data.

À partir de l’exemple de la politique pénitentiaire, la thèse met au jour le rôle du Parlement français dans la réforme d’un secteur d’action publique confiné. Elle montre comment la mobilisation de députés et sénateurs sur un dossier contribue à en redéfinir les problématisations, à susciter des changements incrémentaux dans son traitement et à redessiner les contours de son réseau d’intervenants. La prison a longtemps été gérée par des acteurs administratifs producteurs de leurs propres normes. À l’issue d’un long processus de décloisonnement de l’institution carcérale et de l’éclatement d’un scandale médiatique autour des conditions de détention, deux commissions d’enquête parlementaires font intrusion en prison et initient des apprentissages individuels et collectifs sur celle-ci. Une spécialité parlementaire sur la prison émerge et se consolide sur une décennie, au terme de laquelle la prison est susceptible d'être rapatriée dans le domaine de la loi et de la décision politique. / Dealing with the case of carceral policy, the dissertation sheds light on the role played by the French parliament in reforming a circumscribed branch of public policy. It shows how the mobilisation of MPs on a case contributes to defining this latter's problematising anew, raising incremental changed in the way it is handled and outline new contours for the network of actors involved. Prison has long been managed by professional actors who issued their own norms. Subsequent to a longstanding process whereby the penitentiary institution has been desegmented and public scandals have arisen around emprisonment conditions, two parliament committees have so investigated prisons from within as to engage processes of collective and indiivudal learning about those. A parliamentary specialism on prison has arisen and stabilsied over a decade at the end of which prison is likely to be brought back into the remit of law and political decision-making.

abstract: In two independent and thematically connected chapters, I investigate consumers' willingness to pay a price premium in response to product development that entails prosocial attributes (PATs), those that allude to the reduction of negative externalities to benefit society, and to an innovative participatory pricing design called 'Pay-What-You-Want' (PWYW) pricing, a mechanism that relinquishes the determination of payments in exchange for private goods to the consumers themselves partly relying on their prosocial preferences to drive positive payments. First, I propose a novel statistical approach built on the choice based contingent valuation technique to estimate incremental willingness to pay (IWTP) for PATs that accounts for consumer heterogeneity, dependence in the decision making processes, and incentive compatibility. I validate the approach by estimating IWTP for a variety of PATs and contrast the theoretical and managerial benefits of using the proposed approach over extant techniques used in the literature for this purpose. Second, I propose a general and flexible statistical modeling framework for estimating PWYW payments that exceed zero. It relies on the joint estimation of three types of consumer decision processes namely, the consumer propensity to default to an explicit price recommendation, the propensity to pay a least legitimate price, and the payment of a freely-chosen non-zero payment. Of particular interest is the model's ability to account for a wide variety of design constraints such as the setting of price bounds, explicit price recommendations, and the provision of a menu of discrete prices to choose from. I validate the approach by estimating PWYW payments for a variety of products such as music licenses, snacks, and sports tickets. I specifically examine and report the differential impact of three managerially controllable variables namely, 'payment anonymity', 'information on payment recipients' and 'information of product value/quality'. / Dissertation/Thesis / Doctoral Dissertation Business Administration 2016

Background: Innovation and collaboration are deemed popular terms that are widely used and agreed on. It is implied that innovation without collaboration seems to be unlikely (Deichmann et al., 2017, Haanæs et al., 2018, Innov8rs, 2018). However, the meaning behind innovations of disruptive, incremental, or radical character remains vague and presumably differs between practitioners, just as it differs among scholars. Also, the literature only implies how collaboration can be used effectively for innovation, whilst focussing more on what forms of collaborations are existing.   Research questions: How is innovation understood in the CIC? How does the CIC work together in order to realise cross-industry collaborations for innovation?   Purpose: This study aims at investigating the variety of innovation understanding and how it is applied in collaborations, based on the views of practitioners from different companies that are part of the Cross Industry Club.   Method: This research is designed as an exploratory case study and follows a qualitative strategy with abductive reasoning. Data is collected through nine semi-structured interviews with representatives from five different companies.   Conclusion: The study revealed that practitioners have other aspects in mind which define their understanding of innovation compared to scholars. By that, the usage of innovation terminology is rather arbitrary in practice. The influence of innovation frameworks on the understanding of innovation and the collaborative work have been revealed. Innovation in a cross-industry collaboration, in its purpose and characteristics, is similar to the work of communities of practice. Challenges that appear during the emergence of a collaboration like the CIC, and values that affect the initial work towards cross-industry projects, have been identified.

Made available in DSpace on 2014-12-17T15:16:20Z (GMT). No. of bitstreams: 1 NicoleSOC_DISSERT.pdf: 1342792 bytes, checksum: c040154e5911aeafa634f51d3ef0bb57 (MD5) Previous issue date: 2013-11-29 / Conselho Nacional de Desenvolvimento Cient?fico e Tecnol?gico / obesity affects rightly functional capacity diminishing the cardiovascular system efficiency and oxygen uptake (VO2). Field tests, such as, Incremental Shuttle Walking Test (ISWT) and Six Minute Walk Test (6MWT) has been employed as alternative of Cardiopulmonary Exercise Test (CPX), to functional assessing for conditions which transport of oxygen to peripheral is diminished. Nevertheless, the knowing about metabolic variables response in real time and it comparing among different maximal and submaximal tests in obese is absent. Aim: to compare cardiopulmonary, metabolic response during CPX, ISWT and 6MWT and to analyse it influence of adiposity markers in obese. Material e Method: crosssectional, prospective study. Obese included if: (BMI>30Kg/m2; FVC>80%), were assessed as clinical, anthropometric (BMI, body adiposity index-BAI, waist-WC, hip- HC and neck-NC circumferences) and spirometry (forced vital capacity-FVC, Forced expiratory volume-1?second-FEV1, maximal voluntary ventilation-MVV) variables. Obese performed the sequence of tests: CPX, ISWT and 6MWT. Throughout tests was assessed breath-by-breath by telemetry system (Cortex-Biophysik-Metamax3B) variables; oxygen uptake on peak of activity (VO2peak); carbon dioxide production (VCO2); Volume Expiratory (VE); ventilatory equivalents for VO2 (VE/VO2) and CO2 (VE/VCO2); respiratory exchange rate (RER) and perceived effort-Borg6-20). Results: 15 obese (10women) 39.4+10.1years, normal spirometry (%CVF=93.7+9.7) finished all test. They have BMI (43.5+6.6kg/m2) and different as %adiposity (BAI=50.0+10.5% and 48.8+16.9% respectively women and men). Difference of VO2ml/kg/min and %VO2 were finding between CPX (18.6+4.0) and 6MWT (13.2+2.5) but not between ISWT (15.4+2.9). Agreement was found for ISWT and CPX on VO2Peak (3.2ml/kg/min; 95%; IC-3.0 9.4) and %VO2 (16.4%). VCO2(l/min) confirms similarity in production for CPX (2.3+1.0) and ISWT (1.7+0.7) and difference for 6MWT (1.4+0.6). WC explains more the response of CPX and ISWT than other adiposity markers. Adiposity diminishes 3.2% duration of CPX. Conclusion: ISWT promotes similar metabolic and cardiovascular response than CPX in obese. It suggesting that ISWT could be useful and reliable to assess oxygen uptake and functional capacity in obese / a obesidade afeta diretamente a capacidade funcional diminuindo a efici?ncia do sistema cardiovascular e o consumo de oxig?nio (VO2). Testes de campo, tais como, Incremental Shuttle Walking Test (ISWT) e Teste de Caminhada de 6 minutos (TC6M) tem sido empregados como alternativa ao Teste de Esfor?o Cardiopulmonar (TECP), para avalia??o funcional de patologias que levam a diminui??o da transfer?ncia de oxig?nio ? periferia, entretanto, pouco ? conhecido da respostas fisiol?gicas de vari?veis metab?licas e ventilat?rias em obesos durante teste incremental de exerc?cio. Objetivo: analisar e comparar as respostas cardiopulmonares, metab?licas (VO2pico-consumo de oxig?nio-pico e VCO2-produ??o de di?xido de carbono) e de esfor?o percebido nos testes subm?ximos de esfor?o (ISWT e TC6M) com o TECP. Materiais e M?todo: estudo transversal, prospectivo, onde foi avaliada a concord?ncia de dois diferentes testes de campo (TC6M e ISWT) com o TECP. Crit?rios de Inclus?o (IMC > 30Kg/m2; CVF > 80%). Os participantes realizaram avalia??o cl?nica, antropom?trica (IMC, IAC-?ndice de adiposidade corporal, RCQ-rela??o cintura-quadril, CQ-circunfer?ncia quadril, CC-circunfer?ncia cintura, CP-circunfer?ncia pesco?o) e espirom?trica (CVF-capacidade vital for?ada, VEF1-volume expirat?rio for?ado no 1? segundo, VVM-ventila??o volunt?ria m?xima). Obesos realizaram em momentos distintos a sequ?ncia de testes (TECP, TC6M e ISWT). Medidas ventilat?rias (VE-ventila??o por minuto, VE/VO2-equivalente ventilat?rio de oxig?nio, VE/VCO2-equivalente ventilat?rio de di?xido de carbono, RER-raz?o de troca gasosa) e metab?licas (VO2 e VCO2) dos gases expirados (breath-by-breath) com sistema de telemetria (Cortex-Biophysik-Metamax3B), al?m das vari?veis de percep??o de esfor?o (fadiga e dispneia-Borg6-20) foram tomadas. Resultados: os participantes eram adultos jovens (39.4+10.1 anos), com altos percentuais de adiposidade corporal (IAC=50.0+10.5%-mulheres; IAC=48.8+16.9%- homens) e IMC=43.5+6.6. Observou-se que IMC e CC justificaram isoladamente 56% e 48% da vari?ncia da dura??o do TECP. Diferen?as de VO2ml/kg/min e %VO2 nos obesos foram encontradas apenas entre TECP e TC6M (%VO2 p=0.008 em mulheres e p=0.01 nos homens), sendo os valores de VO2PICO de 18.6+4.0ml/kg/min no TECP; 15.4+2.9 no ISWT e 13.2+2.5 no TC6M. O Bland-Altman evidenciou concord?ncia entre TECP e o ISWT no VO2PICO (3.2ml/kg/min; 95%; IC -3.0 9.4) e no %VO2 (16.4%; 95%; -23.6 56.4). A an?lise da produ??o de CO2(l/min) xiv confirmou uma maior produ??o ao final do TECP (2.3+1.0), seguido do ISWT (1.7+0.7) e TC6M (1.4+0.6), por?m com diferen?as apenas entre o TECP e TC6M (p<0.01). Conclus?o: apesar de considerado um teste de esfor?o subm?ximo, o ISWT promove respostas metab?licas e cardiovasculares semelhantes ao TECP na popula??o obesa, sugerindo que o ISWT pode ser uma boa op??o para avaliar a capacidade funcional de obesos

In this doctoral thesis, we propose new iterative methods for solving a class of convex optimization problems. In general, we consider problems in which the objective function is composed of a finite sum of convex functions and the set of constraints is, at least, convex and closed. The iterative methods we propose are basically designed through the combination of incremental subgradient methods and string-averaging algorithms. Furthermore, in order to obtain methods able to solve optimization problems with many constraints (and possibly in high dimensions), generally given by convex functions, our analysis includes an operator that calculates approximate projections onto the feasible set, instead of the Euclidean projection. This feature is employed in the two methods we propose; one deterministic and the other stochastic. A convergence analysis is proposed for both methods and numerical experiments are performed in order to verify their applicability, especially in large scale problems. / Nesta tese de doutorado, propomos novos métodos iterativos para a solução de uma classe de problemas de otimização convexa. Em geral, consideramos problemas nos quais a função objetivo é composta por uma soma finita de funções convexas e o conjunto de restrições é, pelo menos, convexo e fechado. Os métodos iterativos que propomos são criados, basicamente, através da junção de métodos de subgradientes incrementais e do algoritmo de média das sequências. Além disso, visando obter métodos flexíveis para soluções de problemas de otimização com muitas restrições (e possivelmente em altas dimensões), dadas em geral por funções convexas, a nossa análise inclui um operador que calcula projeções aproximadas sobre o conjunto viável, no lugar da projeção Euclideana. Essa característica é empregada nos dois métodos que propomos; um determinístico e o outro estocástico. Uma análise de convergência é proposta para ambos os métodos e experimentos numéricos são realizados a fim de verificar a sua aplicabilidade, principalmente em problemas de grande escala.

Os métodos incrementais pertencem a uma classe de métodos iterativos que divide o conjunto de dados em subconjuntos ordenados, e que atualiza a imagem ao processar cada subconjunto (sub-iterações). Isso acelera a convergência das reconstruções, e imagens de qualidade são obtidas em menos iterações. No entanto, a cada sub-iteração é necessário calcular os operadores de projeção e retroprojeção, resultando no custo computacional de ordem O(n3) para a reconstrução de imagens de dimensão × . Por outro lado, algumas alternativas baseadas na interpolação em uma grade regular no espaço de Fourier ou em transformadas rápidas não-uniformes, dentre outras ideias, foram desenvolvidas a fim de aliviar esse custo computacional. Além disso, diversas abordagens foram bem sucedidas em acelerar o cálculo das iterações de algoritmos clássicos, mas nenhuma havia sido utilizada em conjunto com os métodos incrementais. Neste trabalho é proposta uma nova abordagem em que a técnica de transformada rápida de Fourier não uniforme (NFFT) é utilizada nas sub-iterações de métodos incrementais com o objetivo de efetuar de forma eficiente os cálculos numericamente mais intensos: a projeção e a retroprojeção, resultando em métodos incrementais com complexidade O(n2 log n ). Os métodos propostos são aplicados à tomografia por radiação síncrotron e os resultados da pesquisa mostram um bom desempenho. / Incremental methods belong to a class of iterative methods that divide the data set into ordered subsets, and which update the image when processing each subset (sub-iterations). It accelerates the reconstruction convergence and quality images are obtained in fewer iterations. However, it is necessary to compute the projection and backprojection operators in each sub-iteration, resulting in the computational cost of O(n3) flops for × images. On the other hand, some alternatives based on interpolation over a regular grid on the Fourier space or on nonequispaced fast transforms, among other ideas, were developed in order to alleviate the computational cost. In addition, several approaches substantially speed up the computation of the iterations of classical algorithms, but the incremental methods had not been benefited from these techniques. In this work, a new approach is proposed in which the nonequispaced fast Fourier transform (NFTT) is used in each subiteration of incremental methods in order to perform the numerically intensive calculations efficiently: the projection and backprojection, resulting in incremental methods with complexity O(n2 log n ). The proposed methods are applied to the synchrotron radiation tomography and the results show a good performance.

O limiar de anaerobiose (LA) é definido como a intensidade de exercício físico em que a produção de energia pelo metabolismo aeróbio é suplementada pelo metabolismo anaeróbio. Este índice constitui-se de um delimitador fisiológico de grande importância para o fornecimento de informações concernentes aos principais sistemas biológicos do organismo, os quais estão envolvidos na realização de um exercício físico. O LA é um importante parâmetro de determinação da capacidade aeróbia funcional de um indivíduo. Diversos métodos são usados para estimar o LA durante exercício. Existem métodos invasivos, como a medida repetida da concentração de lactato sanguíneo; e métodos não-invasivos, por meio de análise de variáveis biológicas como medidas contínuas dos gases respiratórios, através da análise de mudança do padrão de resposta das variáveis ventilatórias e metabólicas, e também pela análise da mudança do padrão de resposta da freqüência cardíaca (FC) frente a um exercício físico incremental. O objetivo deste estudo foi comparar e correlacionar o LA determinado por métodos não-invasivos de análise visual gráfica das variáveis ventilatórias e metabólicas, considerado como padrão-ouro neste estudo, e pelo modelo matemático de regressão linear bi-segmentado utilizando o algoritmo de Hinkley, aplicado a série de dados de FC (Hinkley – FC) e da produção de dióxido de carbono ( CO2) (Hinkley – CO2). Metodologia: Treze mulheres jovens (24 ± 2,63 anos) e dezesseis mulheres na pós-menopausa (57 ± 4,79 anos), saudáveis e sedentárias realizaram teste ergoespirométrico continuo do tipo rampa em cicloergômetro (Quinton Corival 400), com incrementos de 10 a 20 Watts/min até a exaustão física. As variáveis ventilatórias e metabólicas foram captadas respiração a respiração (CPX-D, Medical Graphics), e a FC batimento a batimento (ECAFIX, ACTIVE-E). Os dados foram analisados por testes não paramétricos de Friedman, Mann-Whitney e correlação de Spearman. Nível de significância de ? = 5%. Resultados: Os valores das variáveis potência (W), FC (bpm), consumo de oxigênio relativo ( O2) (mL/kg/min), O2 absoluto (mL/min), CO2 (mL/min) e ventilação pulmonar ( E) (L/min) no LA não apresentaram diferenças significativas entre as metodologias (p > 0,05) nos dois grupos de mulheres estudadas. A análise de correlação dos valores de potência em W, FC em bpm, O2 em mL/kg/min, O2 em mL/min, CO2 em mL/min e E em L/min, entre o método padrão-ouro com o Hinkley – CO2 foram respectivamente: rs=0,75; rs=0,57; rs=0,48; rs=0,66; rs=0,47 e rs=0,46 no grupo jovem, e rs=-0,013; rs=0,77; rs=0,88; rs=0,60; rs=0,76 e rs=0,80 no grupo pós-menopausa. Os valores de correlação do método padrão-ouro com Hinkley – FC para as variáveis potência em W, FC em bpm, O2 em mL/kg/min, O2 em mL/min, CO2 em mL/min e E em L/min, obtidas no LA foram respectivamente: rs=0,58; rs=0,42; rs=0,61; rs=0,57; rs=0,33 e rs=0,39 no grupo de jovens, e rs=0,14; rs=0,87; rs=0,76; rs=0,52; rs=0,33 e rs=0,65 no grupo pós-menopausa. O grupo pós-menopausa apresentou melhores valores de correlação em relação ao grupo de jovens, exceto para as variáveis potência e consumo de oxigênio absoluto (mL/min). Este fato pode estar relacionado a uma maior taxa de variação e magnitude das variáveis analisadas em indivíduos jovens em relação aos de meia-idade, sendo, desta forma, obtida melhor adequação do modelo matemático estudado em mulheres de meia idade. Conclusão: O algoritmo matemático de Hinkley proposto para detectar a mudança no padrão de resposta da CO2 e da FC foi eficiente nos indivíduos de meia-idade, portanto, a metodologia matemática utilizada no presente estudo constitui-se de uma ferramenta promissora para detectar o LA em mulheres saudáveis, por ser um método semi-automatizado, não invasivo e objetivo na determinação do LA. / The anaerobic threshold (AT) is defined as the intensity level of physical exercise at which energy production by aerobic metabolism is supplemented by anaerobic metabolism. This index provides a physiologic delimitation of great importance to supply the organism biological systems information involved in physical exercise performance. The AT constitutes a most important determining of an individuals functional aerobic capacity. Several methods are used for estimating the AT during exercise. There are invasive methods that require repeated blood lactate accumulation; and there exist non-invasive methods by biological variables analysis, like continuous respiratory gases determination by analysis of changes in pattern respiratory and metabolic responses, and heart rate (HR) responses too. The aim of the present study was to compare AT obtained by a graphic visual method of ventilatory and metabolic variables, considered by gold standard method in the present study, with the bi-segmental linear regression mathematic model of Hinkley’s algorithm applied in a HR (Hinkley – HR) and carbon dioxide output ( CO2) (Hinkley – CO2) data. Methodology: Thirteen young women, 24 ± 2,63 years old, and sixteen postmenopausal women, 57 ± 4,79 years old, leading healthy and sedentary life style were submitted to an incremental test in a cicloergometer electromagnetic braking (Quinton Corival 400), with 10 to 20 W/min increments up to physical exhaustion. The ventilatory variables were registered breath-to-breath (CPX-D, Medical Graphics) and HR was obtained beat-to-beat (ECAFIX, ACTIVE-E), over real time. The data were analyzed by Friedman’s test and Spearman’s correlation test, with a level of significance set at 5%. Results: The Power output (W), HR (bpm), oxygen uptake ( O2) (mL/kg/min), O2 (mL/min), CO2 (mL/min) and pulmonary ventilation ( E) (L/min) data in AT have showed no significant differences (p > 0,05) between methods to determine AT in both women groups. The correlation analysis of power output in W, HR in bpm, O2 in mL/kg/min, O2 in mL/min, CO2 in mL/min and E in L/min values, determined by gold standard method and by Hinkley – CO2 data were respectively: rs=0,75; rs=0,57; rs=0,48; rs=0,66; rs=0,47 and rs=0,46 in young group, and rs=-0,013; rs=0,77; rs=0,88; rs=0,60; rs=0,76 and rs=0,80 in postmenopausal group. The correlation analysis by gold standard method and Hinkley – FC in AT of power output in W, HR in bpm, O2 in mL/kg/min, O2 in mL/min, CO2 in mL/min and E in L/min data were respectively: rs=0,58; rs=0,42; rs=0,61; rs=0,57; rs=0,33 and rs=0,39 in young group, and rs=0,14; rs=0,87; rs=0,76; rs=0,52; rs=0,33 and rs=0,65 in postmenopausal group. The postmenopausal group presents better correlations values than young group, except in power output and O2 (mL/min) data. This may be related to more variability rate and higher kinetics responses to variables studied in young group in relation to postmenopausal group. Nevertheless, there was obtained better mathematical model adequacy in middle-age women. Conclusion: the Hinkley’s mathematical algorithm proposed to detect the response patterns changes of CO2 and HR variables was efficient to detect AT in health postmenopausal women’s group, therefore, the mathematical methodology used in the present study showed be a promissory tool because this method represent a semi-automatized, non invasive and objective measure of AT determination.

Nous proposons dans cette thèse trois contributions principales aux méthodes d'apprentissage de dictionnaire. La première est un critère de parcimonie par groupes adapté à la NMF lorsque la mesure de distorsion choisie est la divergence d'Itakura-Saito. Dans la plupart des signaux de musique on peut trouver de longs intervalles où seulement une source est active (des soli). Le critère de parcimonie par groupe que nous proposons permet de trouver automatiquement de tels segments et d'apprendre un dictionnaire adapté à chaque source. Ces dictionnaires permettent ensuite d'effectuer la tâche de séparation dans les intervalles où les sources sont mélangés. Ces deux tâches d'identification et de séparation sont effectuées simultanément en une seule passe de l'algorithme que nous proposons. Notre deuxième contribution est un algorithme en ligne pour apprendre le dictionnaire à grande échelle, sur des signaux de plusieurs heures. L'espace mémoire requis par une NMF estimée en ligne est constant alors qu'il croit linéairement avec la taille des signaux fournis dans la version standard, ce qui est impraticable pour des signaux de plus d'une heure. Notre troisième contribution touche à l'interaction avec l'utilisateur. Pour des signaux courts, l'apprentissage aveugle est particulièrement dificile, et l'apport d'information spécifique au signal traité est indispensable. Notre contribution est similaire à l'inpainting et permet de prendre en compte des annotations temps-fréquences. Elle repose sur l'observation que la quasi-totalité du spectrogramme peut etre divisé en régions spécifiquement assignées à chaque source. Nous décrivons une extension de NMF pour prendre en compte cette information et discutons la possibilité d'inférer cette information automatiquement avec des outils d'apprentissage statistique simples. / In this thesis we provide three main contributions to blind source separation methods based on NMF. Our first contribution is a group-sparsity inducing penalty specifically tailored for Itakura-Saito NMF. In many music tracks, there are whole intervals where only one source is active at the same time. The group-sparsity penalty we propose allows to blindly indentify these intervals and learn source specific dictionaries. As a consequence, those learned dictionaries can be used to do source separation in other parts of the track were several sources are active. These two tasks of identification and separation are performed simultaneously in one run of group-sparsity Itakura-Saito NMF. Our second contribution is an online algorithm for Itakura-Saito NMF that allows to learn dictionaries on very large audio tracks. Indeed, the memory complexity of a batch implementation NMF grows linearly with the length of the recordings and becomes prohibitive for signals longer than an hour. In contrast, our online algorithm is able to learn NMF on arbitrarily long signals with limited memory usage. Our third contribution deals user informed NMF. In short mixed signals, blind learning becomes very hard and sparsity do not retrieve interpretable dictionaries. Our contribution is very similar in spirit to inpainting. It relies on the empirical fact that, when observing the spectrogram of a mixture signal, an overwhelming proportion of it consists in regions where only one source is active. We describe an extension of NMF to take into account time-frequency localized information on the absence/presence of each source. We also investigate inferring this information with tools from machine learning.