Cette thèse de doctorat porte sur la compréhension des aspects concurrentiels du marché de l’innovation incrémentale en France en s’appuyant sur une approche empirique. A cette fin, l’analyse porte plus particulièrement sur l’innovation incrémentale et les dynamiques concurrentielles des médicaments appelés en France « similaires ». La littérature anglophone retient les expressions « me-too » ou « follow-on » pour définir les produits pharmaceutiques à faible valeur ajoutée et qui ont des caractéristiques anatomiques, thérapeutiques et chimiques proches des molécules précédemment mises sur le marché. Le premier chapitre vise notamment l’étude sur les parts de marché des médicaments « follow-on » et le lien avec la variable stratégique définie par l’ordre d’arrivée sur le marché. Dans ce chapitre de thèse, on vérifie que les premiers follow-on d’une classe thérapeutique ont un avantage concurrentiel en ce qui concerne la capacité des capturer et de maintenir des parts de marché pendant une longue période. En outre, les parts de marché sont positivement corrélés avec l’habilité de la firme de fixer un prix supérieur au premier médicament de la classe. Cet avantage stratégique des premiers entrants est aussi intrinsèquement lié aux caractéristiques qualitatives des produits puisque nous avons pu constater que les parts de marché des médicaments sont directement corrélés avec le niveau d’innovation du produit. Le deuxième chapitre de la thèse apporte des éléments sur la nature concurrentielle en termes de prix du marché de l’innovation pharmaceutique incrémentale. Deux variables dépendantes mesurant le prix ont été considérées : le prix du produit calculé en coût du traitement journalier et le prix du médicament divisé par la moyenne des prix des médicaments similaires dans la classe. Nous vérifions que les derniers entrants ont une tendance à avoir un prix inférieur aux premières molécules « follow-on ». Cela implique une caractérisation des derniers entrants ayant un moindre pouvoir de négociation avec le régulateur autour du prix, notamment en raison d’une qualité innovatrice intrinsèque plus faible. Le troisième chapitre présente une analyse empirique autour de la confrontation des deux marchés émergents et d’importance majeure pour la régulation des systèmes de santé : le marché des génériques et le marché des médicaments « follow-on ». L’analyse se concentre sur les aspects de l’intensité de la compétition dans le marché des médicaments brevetés similaires et son impact sur la pénétration des versions génériques de ces derniers. Le constat est que l’intensité de la compétition de médicaments similaires est positivement et significativement corrélée avec l’introduction des génériques. Plus de produits « follow-on » entrainerait donc une baisse encore plus importante des prix des génériques des derniers « follow-on » dans la classe thérapeutique. L’approche essentiellement empirique de cette thèse doctorale permet ainsi de mieux comprendre les déterminants et la dynamique de ce marché relativement émergent et qui suscite des nombreux débats au sein de la communauté scientifique. Enfin nous terminons par une brève conclusion générale fondé autour des résultats de cette recherche permettent d’affirmer que dans un marché régulé comme celui qui prévaut en France, l’intensité de la compétition, engendré notamment par l’arrivé sur le marché des médicaments « follow-on », peut avoir des conséquences positives sur les aspects concurrentiels du secteur du médicament...... / The dynamics of pharmaceutical markets have been constantly changing last years. The development of the so called “follow-on” or “me-too” drugs has been in the centre of a major debate concerning the ability of innovation in the health sector. These drugs are characterized by having a minor level of innovation and do not add any therapeutic value in relation to the previous drugs launched in the market. This doctoral dissertation proposes three empirical essays concerning competition aspects in the market of incremental innovation in France. The first chapter focuses on the impact of entry order on “follow-on” drugs competition in the French market between years 2001 and 2007. More precisely, this study examines the effects on market share of first entrants in the follow-on drug market and how this possible competitive advantage changes over time. Our results are coherent with theoretical microeconomic issues concerning the importance of being first. We find evidence that first movers in the follow on drug market have the ability to capture and maintain greater market share for a long period of time. The hierarchical market position of follow on drugs does not seem to be affected by generic drugs emergence. From a dynamic perspective, our analysis shows that market share is positively correlated with the ability of follow on drugs to set prices higher than the average follow-on drug price in a specific therapeutic class (ATC) which means that market power remains considerably important for first movers. Finally we found that the optimum level of innovation to maximize market share is the highest one.The second chapter examines the relationship between entry order of follow-on drugs and their prices on the French pharmaceutical market. We used a representative panel data of 1047 follow-on drug formulations distributed in 118 ATC classes set over 2001-2007. Two measures of prices were used in the econometric specifications: the absolute price and the relative price of the follow-on drugs. The former concerns simply the absolute price of the drug in daily costs and the latter is the price of the drug relative to the average price of the follow-on drugs in the class. Both prices are calculated on the basis of manufacturer prices. These different indicators give us similar results for the impact of entry order on prices but they are differently correlated with market share. Moreover, our study analyses the potential impact of several variables on prices of pharmaceutical incremental innovation such as firm size, innovation and intensity of competition. Our results suggest that big firms have more ability to negotiate higher prices and that the number of follow-on drugs in the class and the emergence of generic competition may help decrease general prices in the ATC class. We have not found any relationship between prices and innovation in the French pharmaceutical market.The third chapter investigates the potential relationship between follow-on drugs dissemination and generic drug market emergence. It explores the structural determinants of off-patented drugs development at the therapeutic class level with a focus on explanatory variables that reflect the intensity of competition amongst similar interchangeable drugs. We found that generic market growth is greater in therapeutic classes where the number of similar drugs is high and the average brand price is low. This could be due to the fact that brand name drugs reduce their prices to keep market share when generic drugs enter the market. We study also the generic to follow-on brand price ratio at the individual drug level and we found that generic prices of later follow-on drugs are closer to the price of the brand name than generics of first follow-on movers. Our results are coherent with the fact that intensity of competition in the follow-on drugs may help reduce prices not only in the patented drug markets but also in the off-patented sector.

A growing body of research is drawing attention to the importance of open innovation and the reasons firms should progressively switch to this paradigm. However, there is still some reluctance to embrace such approach to innovation. This study investigates the main factors impacting the selection of internal and external new product development in a US multinational company belonging to the semiconductor industry. The main factor found to impact the choice of internal and external innovation is the degree of radicalness of NPD projects. Hence, this is used as additional variable to the internal and external nature of projects to build a matrix, capable of describing the main factors managers take into account when deciding on the projects to undertake. Combining the internal or external and incremental or radical nature of NPD projects, for each category, it was possible to highlight the main dimensions determining the projects selection, namely the expected output and profitability, the purpose of the NPD process, the attention received by managers and the risk involved.

Apresentamos dois novos métodos para a solução de problemas de otimização convexa em dois níveis não necessariamente diferenciáveis, i.e., mostramos que as sequências geradas por ambos os métodos convergem para o conjunto ótimo de uma função não suave sujeito a um conjunto que também envolve a minimização de uma função não diferenciável. Ambos os algoritmos dispensam qualquer tipo de resolução de subproblemas ou busca linear durante suas iterações. Ao final, para demonstrar que os métodos são viáveis, resolvemos um problema de reconstrução de imagens tomográficas / We present two new methods for solving bilevel convex optimization problems, where both functions are not necessarily differentiable, i.e., we show that the sequences generated by those methods converge to the optimal set of a nonsmooth function subject to a set that also involves a function minimization. Both algorithms do not require any kind of subproblems resolution or linear search during the iterations. At the end, to prove that our methods are viable, we solve a problem of tomographic image reconstruction

Árvore de componentes é uma representação completa de imagens que utiliza componentes conexos dos conjuntos de níveis de uma imagem e a relação de inclusão entre esses componentes. Essas informações possibilitam diversas aplicações em processamento de imagens e visão computacional, e.g. filtros conexos, segmentação, extração de características entre outras. Aplicações que utilizam árvore de componentes geralmente computam atributos que descrevem os componentes conexos representados pelos nós da árvore. Entre esses atributos estão a área, o perímetro e o número de Euler, que podem ser utilizados diretamente ou indiretamente (para o cálculo de outros atributos). Os \"bit-quads\" são padrões de tamanho 2x2 binários que são agrupados em determinados conjuntos e contados em imagens binárias. Embora o uso de \"bit-quads\" resulte em um método rápido para calcular atributos em imagens binárias, o mesmo não ocorre para o cálculo de atributos dos nós de uma árvore de componentes, porque os padrões contados em um nó podem se repetir nos conjuntos de níveis da imagem e serem contados mais de uma vez. A literatura recente propõe uma adaptação dos bit-quads para o cálculo incremental e eficiente do número de buracos na árvore de componentes. Essa adaptação utiliza o fato de cada nó da árvore de componentes representar um único componente conexo e uma das definições do número de Euler para o cálculo do número de buracos. Embora essa adaptação possa calcular o número de Euler, os outros atributos (área e perímetro) não podem ser computados. Neste trabalho é apresentada uma extensão dessa adaptação de bit-quads que permite a contagem de todos os agrupamentos de bit-quads de maneira incremental e eficiente na árvore de componentes. De forma que o método proposto possa calcular todos os atributos que podem ser obtidos pelos bit-quads (além do número de buracos) em imagens binárias na árvore de componentes de maneira incremental. / Component tree is a full image representation which uses the connected components of the level sets of the image and these connected components\' inclusion relationship. This information can be used in various image processing and computational vision applications, e.g. connected filters, segmentation, feature extraction, among others. In general, applications which use component trees compute attributes that describe the connected components represented by the tree nodes. Attributes such as area, perimeter and Euler number, can be used directly or indirectly (when they are used to compute other attributes) to describe the component tree nodes in various applications. The bit-quads are binary patterns of size 2x2 that are grouped in determined sets and counted in binary images to compute area, perimeter (also their continuous approximation) and Euler number. Even though the bit-quads usage can yield an efficient method to compute binary image attributes, they cannot be used efficiently to compute attributes for all component tree nodes, since some bit-quads can be counted more than once over the level sets. An adaptation of the bit-quads has been proposed to compute efficiently and incrementally the number of holes for all component tree nodes. This adaptation uses the fact that each component tree node represents a unique connected component and one of Euler number definitions to compute the number of holes. Even though this adaptation can compute Euler number, it cannot compute other attributes derived from the bit-quads (area and perimeter). In this work, an extension of this adaptation is proposed to efficiently and incrementally count all bit-quads sets in a component tree. Moreover, it yields a method to compute all attributes which can be computed by the bit-quads in binary images in the component tree using an incremental strategy.

Electricity network pricing approaches play a fundamental role in establishing whether providing the network service function is economically beneficial to both the network operators and other stakeholders, namely, network users. Many pricing methodologies have been developed since the late 80‟s. The earlier approaches were not based on economic principle while the latest are directed to being more based on economic principle as the shift is towards deregulated and privatized electric power industry as opposed to the earlier vertically regulated regime. As a result, many such methodologies based on economic principle have emerged and these reflect the investment cost incurred in circuits and transformers to support real and reactive power flow. However, to reflect investment cost incurred for maintaining network voltages in network charges has received very little attention in network charges. Therefore, this research work is aimed to create a charging approach to recover investment cost, by the network operator, for maintaining the network voltages. This thesis presents a new long-run incremental cost (LRIC) pricing approach for distribution networks and demonstrates the course of action of evaluating and allocating the network asset cost in the context of maintaining network voltages. Also, it should be noted that this approach can be used for transmission networks. Firstly, the LRIC-voltage network pricing approach for reflecting the future network VAr compensation assets is proposed. Then, this approach is extended to consider n-1 contingency situation as per statutory requirement that the network should be able to withstand such contingencies in order to enhance reasonable security and reliability in its network. Lastly, this LRIC-voltage network charging methodology is again extended to reflect the charges for existing network VAr compensation assets. In addition, this LRIC-voltage network pricing approach is improved to reflect better the nodal charges as the respective nodal voltage degradation rates, given corresponding load growth rate, are determined based on the P-V curve concept. The advantages of all these incorporate the ability to reflect correct forward-looking charges, to recognize both real and reactive powers, to provide locational charges and to provide charges for both generation and demand customers. In addition, two fundamental studies were conducted to demonstrate the trend in which the LRIC-voltage network charges would follow given different networks and different load growth rates. What set apart the LRIC-voltage network charges are those two parameters. Moreover, with regard to different networks, this was a defining moment as to how the aforementioned charges should be sought given transmission and distribution networks. A pricing software package utilizing load-flow has been developed implementing the proposed LRIC-voltage network pricing methodology and, its extensions. This software can well be utilized by transmission and distribution companies for analyzing their cost. The LRIC-voltage network pricing methodology and its extensions, are all demonstrated on the IEEE 14-bus test system and a practical distribution test network in the South Wales area of England, UK.

The purpose of this study was to predict VO2pak using ratings of perceived exertion (RPE), heart rate (HR), and percent fat (PFAT). Subjects were males (n= 60) (PFAT, M SD = 14.4 6.1) and females (n= 67) (PFAT, M SD = 23.4 4.9) with ages ranging from 18 to 33 years. Subjects performed an incremental cycle ergometer protocol and RPE, HR and Vo2 were measured at each stage until VO2 ak was achieved. Mean RPE and HR at the submaximal workload of 100 watts were, (RPE100) M= 12.7 2.6 and (HR100) M= 146.924.7 respectively. Correlations (p< .001) with VO2p. were -.75 (PFAT), -.66 (HR100), -.67 (FIPE100). The multiple correlation using PFAT, HR100, and RPE100 as predictors of VO2pak was .83 (SEE= 5.28 ml-kg BW'smin"). Each predictor contributed to the correlation (p<.01). The results indicate that PFAT combined with exercise responses of RPE and HR provide valid estimates of VO2peak with a relatively small SEE.

"Keeping views fresh by maintaining the consistency between materialized views and their base data in the presence of base updates is a critical problem for many applications, including data warehousing and data integration. While heavily studied for traditional databases, the maintenance of XML views remains largely unexplored. Maintaining XML views is complex due to the richness of the XML data model and the powerful capabilities of XML query languages, such as XQuery. This dissertation proposes a comprehensive solution for the general problem of maintaining materialized XQuery views. Our solution is the first to enable the maintenance of a large class of XQuery views including XPath expressions, FLWOR expressions, and Element Constructors. These views may contain arbitrary result construction and arbitrary grouping and join operations. Our solution also supports the unique order requirements of XQuery including source document order and query order. The contributions of this dissertation include: (i) an efficient solution for supporting order in XML query processing and view maintenance, (ii) an identifier-based technique for enabling incremental construction of XML views, (iii) a mechanism for modeling and validating source XML updates, (iv) a counting algorithm for supporting view maintenance on delete and modify updates, (v) an algebraic solution for propagating bulk XML updates, and (vi) an efficient mechanism for refreshing materialized XML views on propagated updates. We provide proofs of correctness of our proposed techniques for materialized XQuery maintenance. We have implemented a prototype of our view maintenance solution on top of the Rainbow XML query engine, developed at WPI. Our experiments confirm that our solution provides a practical and efficient solution for maintaining materialized XQuery views even when handling heterogeneous batches of possibly large source updates. Our solution follows the widely adopted propagate-apply framework for view maintenance common to all mainstream query engines. That is, our solution produces incremental maintenance plans in the same algebraic language used to define the views. These plans can thus be optimized and executed by standard query processing techniques. Being compatible with standard frameworks paves the way for our XML view maintenance solution to be easily adopted by existing database engines."

The ability to organize concepts (e.g., dog, chair) into efficient mental representations, i.e., categories (e.g., animal, furniture) is a fundamental mechanism which allows humans to perceive, organize, and adapt to their world. Much research has been dedicated to the questions of how categories emerge and how they are represented. Experimental evidence suggests that (i) concepts and categories are represented through sets of features (e.g., dogs bark, chairs are made of wood) which are structured into different types (e.g, behavior, material); (ii) categories and their featural representations are learnt jointly and incrementally; and (iii) categories are dynamic and their representations adapt to changing environments. This thesis investigates the mechanisms underlying the incremental and dynamic formation of categories and their featural representations through cognitively motivated Bayesian computational models. Models of category acquisition have been extensively studied in cognitive science and primarily tested on perceptual abstractions or artificial stimuli. In this thesis, we focus on categories acquired from natural language stimuli, using nouns as a stand-in for their reference concepts, and their linguistic contexts as a representation of the concepts’ features. The use of text corpora allows us to (i) develop large-scale unsupervised models thus simulating human learning, and (ii) model child category acquisition, leveraging the linguistic input available to children in the form of transcribed child-directed language. In the first part of this thesis we investigate the incremental process of category acquisition. We present a Bayesian model and an incremental learning algorithm which sequentially integrates newly observed data. We evaluate our model output against gold standard categories (elicited experimentally from human participants), and show that high-quality categories are learnt both from child-directed data and from large, thematically unrestricted text corpora. We find that the model performs well even under constrained memory resources, resembling human cognitive limitations. While lists of representative features for categories emerge from this model, they are neither structured nor jointly optimized with the categories. We address these shortcomings in the second part of the thesis, and present a Bayesian model which jointly learns categories and structured featural representations. We present both batch and incremental learning algorithms, and demonstrate the model’s effectiveness on both encyclopedic and child-directed data. We show that high-quality categories and features emerge in the joint learning process, and that the structured features are intuitively interpretable through human plausibility judgment evaluation. In the third part of the thesis we turn to the dynamic nature of meaning: categories and their featural representations change over time, e.g., children distinguish some types of features (such as size and shade) less clearly than adults, and word meanings adapt to our ever changing environment and its structure. We present a dynamic Bayesian model of meaning change, which infers time-specific concept representations as a set of feature types and their prevalence, and captures their development as a smooth process. We analyze the development of concept representations in their complexity over time from child-directed data, and show that our model captures established patterns of child concept learning. We also apply our model to diachronic change of word meaning, modeling how word senses change internally and in prevalence over centuries. The contributions of this thesis are threefold. Firstly, we show that a variety of experimental results on the acquisition and representation of categories can be captured with computational models within the framework of Bayesian modeling. Secondly, we show that natural language text is an appropriate source of information for modeling categorization-related phenomena suggesting that the environmental structure that drives category formation is encoded in this data. Thirdly, we show that the experimental findings hold on a larger scale. Our models are trained and tested on a larger set of concepts and categories than is common in behavioral experiments and the categories and featural representations they can learn from linguistic text are in principle unrestricted.

Recentemente, houve um aumento rápido na criação e disponibilidade de repositórios de dados, o que foi percebido nas áreas de Mineração de Dados e Aprendizagem de Máquina. Este fato deve-se principalmente à rápida criação de tais dados em redes sociais. Uma grande parte destes dados é feita de texto, e a informação armazenada neles pode descrever desde perfis de usuários a temas comuns em documentos como política, esportes e ciência, informação bastante útil para várias aplicações. Como muitos destes dados são criados em fluxos, é desejável a criação de algoritmos com capacidade de atuar em grande escala e também de forma on-line, já que tarefas como organização e exploração de grandes coleções de dados seriam beneficiadas por eles. Nesta dissertação um modelo probabilístico, on-line e incremental é apresentado, como um esforço em resolver o problema apresentado. O algoritmo possui o nome DV-INBC e é uma extensão ao algoritmo INBC. As duas principais características do DV-INBC são: a necessidade de apenas uma iteração pelos dados de treino para criar um modelo que os represente; não é necessário saber o vocabulário dos dados a priori. Logo, pouco conhecimento sobre o fluxo de dados é necessário. Para avaliar a performance do algoritmo, são apresentados testes usando datasets populares. / Recently the fields of Data Mining and Machine Learning have seen a rapid increase in the creation and availability of data repositories. This is mainly due to its rapid creation in social networks. Also, a large part of those data is made of text documents. The information stored in such texts can range from a description of a user profile to common textual topics such as politics, sports and science, information very useful for many applications. Besides, since many of this data are created in streams, scalable and on-line algorithms are desired, because tasks like organization and exploration of large document collections would be benefited by them. In this thesis an incremental, on-line and probabilistic model for document classification is presented, as an effort of tackling this problem. The algorithm is called DV-INBC and is an extension to the INBC algorithm. The two main characteristics of DV-INBC are: only a single scan over the data is necessary to create a model of it; the data vocabulary need not to be known a priori. Therefore, little knowledge about the data stream is needed. To assess its performance, tests using well known datasets are presented.

Classificação de fluxos contínuos de dados possui muitos desafios para a comunidade de mineração de dados quando o ambiente não é estacionário. Um dos maiores desafios para a aprendizagem em fluxos contínuos de dados está relacionado com a adaptação às mudanças de conceito, as quais ocorrem como resultado da evolução dos dados ao longo do tempo. Duas formas principais de desenvolver abordagens adaptativas são os métodos baseados em conjunto de classificadores e os algoritmos incrementais. Métodos baseados em conjunto de classificadores desempenham um papel importante devido à sua modularidade, o que proporciona uma maneira natural de se adaptar a mudanças de conceito. Os algoritmos incrementais são mais rápidos e possuem uma melhor capacidade anti-ruído do que os conjuntos de classificadores, mas têm mais restrições sobre os fluxos de dados. Assim, é um desafio combinar a flexibilidade e a adaptação de um conjunto de classificadores na presença de mudança de conceito, com a simplicidade de uso encontrada em um único classificador com aprendizado incremental. Com essa motivação, nesta dissertação, propomos um algoritmo incremental, online e probabilístico para a classificação em problemas que envolvem mudança de conceito. O algoritmo é chamado IGMN-NSE e é uma adaptação do algoritmo IGMN. As duas principais contribuições da IGMN-NSE em relação à IGMN são: melhoria de poder preditivo para tarefas de classificação e a adaptação para alcançar um bom desempenho em cenários não estacionários. Estudos extensivos em bases de dados sintéticas e do mundo real demonstram que o algoritmo proposto pode rastrear os ambientes em mudança de forma muito próxima, independentemente do tipo de mudança de conceito. / Data stream classification poses many challenges for the data mining community when the environment is non-stationary. The greatest challenge in learning classifiers from data stream relates to adaptation to the concept drifts, which occur as a result of changes in the underlying concepts. Two main ways to develop adaptive approaches are ensemble methods and incremental algorithms. Ensemble method plays an important role due to its modularity, which provides a natural way of adapting to change. Incremental algorithms are faster and have better anti-noise capacity than ensemble algorithms, but have more restrictions on concept drifting data streams. Thus, it is a challenge to combine the flexibility and adaptation of an ensemble classifier in the presence of concept drift, with the simplicity of use found in a single classifier with incremental learning. With this motivation, in this dissertation we propose an incremental, online and probabilistic algorithm for classification as an effort of tackling concept drifting. The algorithm is called IGMN-NSE and is an adaptation of the IGMN algorithm. The two main contributions of IGMN-NSE in relation to the IGMN are: predictive power improvement for classification tasks and adaptation to achieve a good performance in non-stationary environments. Extensive studies on both synthetic and real-world data demonstrate that the proposed algorithm can track the changing environments very closely, regardless of the type of concept drift.