The necessity of modeling is vital for the purpose of reasoning and diagnosing in complex systems, since the human mind might sometimes have a limited capacity and an inability to be objective. The chain graph (CG) class is a powerful and robust tool for modeling real-world applications. It is a type of probabilistic graphical models (PGM) and has multiple interpretations. Each of these interpretations has a distinct Markov property. This thesis deals with the multivariate regression chain graph (MVR-CG) interpretation. The main goal of this thesis is to implement and evaluate the results of the MVR-PC-algorithm proposed by Sonntag and Peña in 2012. This algorithm uses a constraint based approach used in order to learn a MVR-CG from data.In this study the MRV-PC-algorithm is implemented and tested to see whether the implementation is correct. For this purpose, it is run on several different independence models that can be perfectly represented by MVR-CGs. The learned CG and the independence model of the given probability distribution are then compared to ensure that they are in the same Markov equivalence class. Additionally, for the purpose of checking how accurate the algorithm is, in learning a MVR-CG from data, a large number of samples are passed to the algorithm. The results are analyzed based on number of nodes and average number of adjacents per node. The accuracy of the algorithm is measured by the precision and recall of independencies and dependencies.In general, the higher the number of samples given to the algorithm, the more accurate the learned MVR-CGs become. In addition, when the graph is sparse, the result becomes significantly more accurate. The number of nodes can affect the results slightly. When the number of nodes increases it can lead to better results, if the average number of adjacents is fixed. On the other hand, if the number of nodes is fixed and the average number of adjacents increases, the effect is more considerable and the accuracy of the results dramatically declines. Moreover the type of the random variables can affect the results. Given the samples with discrete variables, the recall of independencies measure would be higher and the precision of independencies measure would be lower. Conversely, given the samples with continuous variables, the recall of independencies would be less but the precision of independencies would be higher.

En Guadeloupe, l'utilisation de composts se heurte à de nombreux freins, aussi bien en termes de leur qualité qu’en termes d’un manque de plateformes de compostage à grande échelle et de proximité. Des études récentes ont montré que la qualité des composts locaux était plus faible qu’en milieu tempéré, ce qui constituerait un verrou majeur à l’adoption de la pratique et l’utilisation des composts industriels locaux. Pourtant, les objectifs de valorisation des déchets organiques fixés par les instances publiques sont ambitieux et le gisement local, bien que diffus et actuellement mal géré ou négligé, présenterait un réel intérêt pour la profession agricole à être orienté vers la valorisation biologique telle que le compostage. Dans ce contexte, l’objectif de ce travail était d'évaluer la qualité des composts élaborés en milieu tropical et d'apporter des éléments factuels pour son amélioration et, plus spécifiquement, d’apporter de l’information sur les potentialités agronomiques du compostage des effluents d’élevage en Guadeloupe, présentant ainsi le compostage comme une alternative durable pour le recyclage des déchets d’origine animale.Une méta-analyse de 442 composts d'origine diverse, la première réalisée sur le sujet, nous a permis de démontrer que les composts produits en milieu tropical présentent des teneurs en carbone, azote, potassium et fraction soluble de la matière organique plus faibles que celles des composts produits en milieu tempéré, et que cela pourrait notamment être dû à l’influence des conditions climatiques lors du compostage. En revanche, nous avons pu mettre en évidence que certaines matières premières permettaient l’obtention de composts de meilleure qualité quelque soit le climat considéré, il s’agissait entre autres des effluents d’élevage.Les résultats issus d'une série d’expérimentations menée sur la production de composts d’effluents d’élevage avec co-compostage et vermicompostage ont été traités avec une approche méthodologique innovante dans ce domaine, les Réseaux Bayésiens. L’évaluation réalisée sur le co-compostage effluents/déchets verts nous a permis d’identifier l’"effet de concentration" du carbone et de la lignine, comme celui qui définit la qualité des composts en termes de quantité et de stabilité de la matière organique. En revanche, dans le cas des nutriments, seule la qualité des matières premières a été identifiée comme le facteur déterminant de la qualité des produits finaux. Ces résultats nous ont amené à considérer les effluents d’élevage de bovin comme la matière première la plus efficace pour produire des co-composts de qualité satisfaisante, répondant à la problématique d’usure de la matière organique des sols guadeloupéens et permettant de satisfaire les attentes de la profession agricole.De plus des expérimentations réalisées sur les composts domestiques ont montré que la gamme analysée présentait une variabilité trop importante pour être considérée comme acceptable par la profession agricole. Le compostage domestique peut permettre de produire des composts de bonne qualité agronomique à utiliser à la petite échelle des jardins particuliers et des jardins créoles. / In Guadeloupe, the practice of composting faces many obstacles and preconceptions both in terms of quality and in terms of lack of large-scale composting plants as well as local composting facilities. Recent studies have shown that the quality of local composts was lower compared to those from temperate regions. This constitutes an important constraint for the adoption of the former by farmers. However, organic waste recovery targets set by the government are ambitious and local resources, although diffused and currently poorly managed or neglected would be of real interest for the farming profession by being directed towards organic recycling such as composting. In this context, the aim of this thesis was to evaluate the quality of compost produced in the tropics, provide factual elements for improvement and, more specifically, to provide information on the agronomic potential of composting livestock manure in Guadeloupe, presenting composting as a sustainable alternative for the recycling of animal waste.A meta-analysis of 442 composts from various sources, the first one to be conducted on the subject enabled us to demonstrate that composts produced in the tropics present lower contents of carbon, nitrogen, potassium and soluble fraction of organic matter than those produced in temperate environments. This could especially be due to the influence of climatic conditions during composting. However, we were able to show that some raw materials allow better quality composts whatever the considered climate, especially the case of livestock manure.A series of experiments conducted on the production of livestock manure composts with co-composting and vermicomposting were treated with an innovative methodological approach in this field, the Bayesian Networks. The evaluation carried out on co-composting has allowed us to identify that the "concentration effect" was the main factor affecting compost quality in terms of amount and stability of organic matter. While in the case of nutrients, only the quality of raw materials has been identified as the determining factor affecting the quality of the end products. These results led us to consider manure, mainly cattle manure, as the most efficient feedstock for producing satisfactory quality composts, meeting the needs of loss of soil organic matter in Guadeloupe and the needs of the farming profession.Further experiments performed on household composts showed that their quality exhibited a too important variability to be considered acceptable by farmers. Our results indicate that household composts could be suitable for use in small-scale private gardens and Creole gardens.

Cette thèse explore l’utilisation des réseaux Bayésiens (RB) afin de répondre à des problématiques de dégradation en grandes dimensions concernant des infrastructures du génie civil. Alors que les approches traditionnelles basées l’évolution physique déterministe de détérioration sont déficientes pour des problèmes à grande échelle, les gestionnaires d’ouvrages ont développé une connaissance de modèles nécessitant la gestion de l’incertain. L’utilisation de la dépendance probabiliste se révèle être une approche adéquate dans ce contexte tandis que la possibilité de modéliser l’incertain est une composante attrayante. Le concept de dépendance au sein des RB s’exprime principalement de deux façons. D’une part, les probabilités conditionnelles classiques s’appuyant le théorème de Bayes et d’autre part, une classe de RB faisant l’usage de copules et corrélation de rang comme mesures de dépendance. Nous présentons à la fois des contributions théoriques et pratiques dans le cadre de ces deux classes de RB ; les RB dynamiques discrets et les RB non paramétriques, respectivement. Des problématiques concernant la paramétrisation de chacune des classes sont également abordées. Dans un contexte théorique, nous montrons que les RBNP permet de caractériser n’importe quel processus de Markov. / This thesis explores high-dimensional deterioration-related problems using Bayesian networks (BN). Asset managers become more and more familiar on how to reason with uncertainty as traditional physics-based models fail to fully encompass the dynamics of large-scale degradation issues. Probabilistic dependence is able to achieve this while the ability to incorporate randomness is enticing.In fact, dependence in BN is mainly expressed in two ways. On the one hand, classic conditional probabilities that lean on thewell-known Bayes rule and, on the other hand, a more recent classof BN featuring copulae and rank correlation as dependence metrics. Both theoretical and practical contributions are presented for the two classes of BN referred to as discrete dynamic andnon-parametric BN, respectively. Issues related to the parametrization for each class of BN are addressed. For the discrete dynamic class, we extend the current framework by incorporating an additional dimension. We observed that this dimension allows to have more control on the deterioration mechanism through the main endogenous governing variables impacting it. For the non-parametric class, we demonstrate its remarkable capacity to handle a high-dimension crack growth issue for a steel bridge. We further show that this type of BN can characterize any Markov process.

Les fautes systèmes peuvent conduire à des conséquences sérieuses pour l’humain, l’environnement et le matériel. Or, y remédier peut s’avérer coûteux voire même dangereux. Ainsi, afin d’éviter ces situations, il est devenu essentiel pour les systèmes complexes modernes de détecter et d’identifier tout changement dans leur fonctionnement nominal avant que cela ne devienne critique. De ce fait, plusieurs méthodes de détection et de diagnostic ont été proposées ou améliorées durant les dernières décennies. Parmi ces méthodes, celles présentant un fort intérêt se basent sur un outil statistique et probabiliste nommé réseau Bayésien. Toutefois, la majorité d’entre elles ne tiennent pas compte du risque de fausse alarme dans leur prise de décision. L’intérêt de cette thèse est alors d’introduire sous réseau Bayésien des limites probabilistes permettant le respect d’un niveau de signification considéré. Plus exactement, nous proposons une modélisation des statistiques quadratiques et les limites leurs correspondant sur réseau Bayésien. Ceci nous permet de généraliser sous réseau Bayésien des schémas de détection de fautes comme par exemple ceux basés sur l’analyse en composantes principale. Cette modélisation nous permet également de proposer une famille de réseaux Bayésiens permettant de faire de la détection et du diagnostic de façon simultanée, tout en tenant compte d’un rejet de distance. Enfin, nous proposons un cadre probabiliste permettant d’unifier les différents réseaux Bayésiens pouvant être utilisés pour la détection ou le diagnostic de fautes. / Systems failures can potentially lead to serious consequences forhuman, environment and material, and sometimes fixing them could be expensive and even dangerous. Thus, in order to avoid these undesirable situations, it becomes very important and essential for modern complex systems to detect and identify any changes in their nominal operations before they become critical. To do so, several detection and diagnosis methods have been proposed or enhanced during the last decades. Among these methods, those with a great interest are based on a statistical and probabilistic tool named Bayesian network. However, the majority of these methods do not handle the risk of false alarm in their decision-making. The interest of this thesis is to introduce, under Bayesian network, probabilistic limits able to respect a given significance level. More precisely, we propose to model the quadratic statistics and their limits in Bayesian network. This allows us to generalize under Bayesian network fault detection schemes as those associated to the principal component analysis. This modeling allows us also to propose a family of Bayesian networks that can make detection and diagnosis simultaneously, while taking into account the distance rejection.Finally, we propose a probabilistic framework able to unify different BNs dedicated to the detection or diagnosis of systems faults.

Les systèmes de dialogue homme-machine ont pour objectif de permettre un échange oral efficace et convivial entre un utilisateur humain et un ordinateur. Leurs domaines d'applications sont variés, depuis la gestion d'échanges commerciaux jusqu'au tutorat ou l'aide à la personne. Cependant, les capacités de communication de ces systèmes sont actuellement limités par leur aptitude à comprendre la parole spontanée. Nos travaux s'intéressent au module de compréhension de la parole et présentent une proposition entièrement basée sur des approches stochastiques, permettant l'élaboration d'une hypothèse sémantique complète. Notre démarche s'appuie sur une représentation hiérarchisée du sens d'une phrase à base de frames sémantiques. La première partie du travail a consisté en l'élaboration d'une base de connaissances sémantiques adaptée au domaine du corpus d'expérimentation MEDIA (information touristique et réservation d'hôtel). Nous avons eu recours au formalisme FrameNet pour assurer une généricité maximale à notre représentation sémantique. Le développement d'un système à base de règles et d'inférences logiques nous a ensuite permis d'annoter automatiquement le corpus. La seconde partie concerne l'étude du module de composition sémantique lui-même. En nous appuyant sur une première étape d'interprétation littérale produisant des unités conceptuelles de base (non reliées), nous proposons de générer des fragments sémantiques (sous-arbres) à l'aide de réseaux bayésiens dynamiques. Les fragments sémantiques générés fournissent une représentation sémantique partielle du message de l'utilisateur. Pour parvenir à la représentation sémantique globale complète, nous proposons et évaluons un algorithme de composition d'arbres décliné selon deux variantes. La première est basée sur une heuristique visant à construire un arbre de taille et de poids minimum. La seconde s'appuie sur une méthode de classification à base de séparateurs à vaste marge pour décider des opérations de composition à réaliser. Le module de compréhension construit au cours de ce travail peut être adapté au traitement de tout type de dialogue. Il repose sur une représentation sémantique riche et les modèles utilisés permettent de fournir des listes d'hypothèses sémantiques scorées. Les résultats obtenus sur les données expérimentales confirment la robustesse de l'approche proposée aux données incertaines et son aptitude à produire une représentation sémantique consistante / Spoken dialog systems enable users to interact with computer systems via natural dialogs, as they would with human beings. These systems are deployed into a wide range of application fields from commercial services to tutorial or information services. However, the communication skills of such systems are bounded by their spoken language understanding abilities. Our work focus on the spoken language understanding module which links the automatic speech recognition module and the dialog manager. From the user’s utterance analysis, the spoken language understanding module derives a representation of its semantic content upon which the dialog manager can decide the next best action to perform. The system we propose introduces a stochastic approach based on Dynamic Bayesian Networks (DBNs) for spoken language understanding. DBN-based models allow to infer and then to compose semantic frame-based tree structures from speech transcriptions. First, we developed a semantic knowledge source covering the domain of our experimental corpus (MEDIA, a French corpus for tourism information and hotel booking). The semantic frames were designed according to the FrameNet paradigm and a hand-craft rule-based approach was used to derive the seed annotated training data.Then, to derive automatically the frame meaning representations, we propose a system based on a two decoding step process using DBNs : first basic concepts are derived from the user’s utterance transcriptions, then inferences are made on sequential semantic frame structures, considering all the available previous annotation levels. The inference process extracts all possible sub-trees according to lower level information and composes the hypothesized branches into a single utterance-span tree. The composition step investigates two different algorithms : a heuristic minimizing the size and the weight of the tree ; a context-sensitive decision process based on support vector machines for detecting the relations between the hypothesized frames. This work investigates a stochastic process for generating and composing semantic frames using DBNs. The proposed approach offers a convenient way to automatically derive semantic annotations of speech utterances based on a complete frame hierarchical structure. Experimental results, obtained on the MEDIA dialog corpus, show that the system is able to supply the dialog manager with a rich and thorough representation of the user’s request semantics

The objective of the research presented in this thesis is to investigate alternative ways of information processing employing asynchronous, data driven, and analogue computation in massively parallel cellular processor arrays, with applications in machine vision and artificial intelligence. The use of cellular processor architectures, with only local neighbourhood connectivity, is considered in VLSI realisations of the trigger-wave propagation in binary image processing, and in Bayesian inference. Design issues, critical in terms of the computational precision and system performance, are extensively analysed, accounting for the non-ideal operation of MOS devices caused by the second order effects, noise and parameter mismatch. In particular, CMOS hardware solutions for two specific tasks: binary image skeletonization and sum-product algorithm for belief propagation in factor graphs, are considered, targeting efficient design in terms of the processing speed, power, area, and computational precision. The major contributions of this research are in the area of continuous-time and discrete-time CMOS circuit design, with applications in moderate precision analogue and asynchronous computation, accounting for parameter variability. Various analogue and digital circuit realisations, operating in the continuous-time and discrete-time domains, are analysed in theory and verified using combined Matlab-Hspice simulations, providing a versatile framework suitable for custom specific analyses, verification and optimisation of the designed systems. Novel solutions, exhibiting reduced impact of parameter variability on the circuit operation, are presented and applied in the designs of the arithmetic circuits for matrix-vector operations and in the data driven asynchronous processor arrays for binary image processing. Several mismatch optimisation techniques are demonstrated, based on the use of switched-current approach in the design of current-mode Gilbert multiplier circuit, novel biasing scheme in the design of tunable delay gates, and averaging technique applied to the analogue continuous-time circuits realisations of Bayesian networks. The most promising circuit solutions were implemented on the PPATC test chip, fabricated in a standard 90 nm CMOS process, and verified in experiments.

A extração de Interações Proteína-Proteína (IPPs) a partir de texto é um problema relevante na área biomédica e um desafio na área de aprendizado de máquina. Na área biomédica, as IPPs são fundamentais para compreender o funcionamento dos seres vivos. No entanto, o número de artigos relacionados com IPPs está aumentando rapidamente, sendo impraticável identicá-las e catalogá-las manualmente. Por exemplo, no caso das IPPs humanas apenas 10% foram catalogadas. Por outro lado, em aprendizado de máquina, métodos baseados em kernels são frequentemente empregados para extrair automaticamente IPPs, atingindo resultados considerados estado da arte. Esses métodos usam informações léxicas, sintáticas ou semânticas como características. Entretanto, os resultados ainda são insuficientes, atingindo uma taxa relativamente baixa, em termos da medida F, devido à complexidade do problema. Apesar dos esforços em produzir kernels, cada vez mais sofisticados, usando árvores sintáticas como árvores constituintes ou de dependência, pouco é conhecido sobre o desempenho de outras abordagens de aprendizado de máquina como, por exemplo, as redes bayesianas. As àrvores constituintes são estruturas de grafos que contêm informação importante da gramática subjacente as sentenças de textos contendo IPPs. Por outro lado, a rede bayesiana permite modelar algumas regras da gramática e atribuir para elas uma distribuição de probabilidade de acordo com as sentenças de treinamento. Neste trabalho de mestrado propõe-se um método para construção automática de redes bayesianas a partir de árvores contituintes para extração de IPPs. O método foi testado em cinco corpora padrões da extração de IPPs, atingindo resultados competitivos, em alguns casos melhores, em comparação a métodos do estado da arte / Extracting Protein-Protein Interactions (PPIs) from text is a relevant problem in the biomedical field and a challenge in the area of machine learning. In the biomedical field, the PPIs are fundamental to understand the functioning of living organisms. However, the number of articles related to PPIs is increasing rapidly, hence it is impractical to identify and catalog them manually. For example, in the case of human PPIs only 10 % have been cataloged. On the other hand, machine learning methods based on kernels are often employed to automatically extract PPIs, achieving state of the art results. These methods use lexical, syntactic and semantic information as features. However, the results are still poor, reaching a relatively low rate of F-measure due to the complexity of the problem. Despite efforts to produce sophisticate kernels, using syntactic trees as constituent or dependency trees, little is known about the performance of other Machine Learning approaches, eg, Bayesian networks. Constituent tree structures are graphs which contain important information of the underlying grammar in sentences containing PPIs. On the other hand, the Bayesian network allows modeling some rules of grammar and assign to them a probability distribution according to the training sentences. In this master thesis we propose a method for automatic construction of Bayesian networks from constituent trees for extracting PPIs. The method was tested in five corpora, considered benchmark of extraction of PPI, achieving competitive results, and in some cases better results when compared to state of the art methods

Nesta tese são apresentados um novo modelo estatístico para cálculo de probabilidade de paternidade e sua implementação em software. O modelo proposto utiliza o genótipo como informação básica, em contraste com outros modelos que usam alelos. Por esta diferença, o modelo proposto resulta mais abrangente, mas que, sob certas restrições, reproduz os resultados dos modelos que usam alelos. Este modelo foi implementado em um software que recebe descrições da genealogia e dos marcadores em uma linguagem dedicada a isso e constrói uma rede bayesiana para cada marcador. O usuário pode definir livremente a genealogia e os marcadores. O cálculo da probabilidade de paternidade é feito, sobre as redes construídas, por um software para inferência em redes bayesianas e a probabilidade de paternidade combinada considerando todos os marcadores é calculada, resultando em um \"índice de paternidade. / This thesis presents a novel statistical model for calculation of the probability of paternity and its implementation as a software. The proposed model uses genotype as basic information. Other models use alleles as basic information. As a result the proposed model is broader, in the sense that, under certain constraints the results from the other models are reproduced. The software implementation receives pedigree and markers data, in a specifically designed language, as input and builds one bayesian network for each marker. The user can freely define any pedigree and any marker. Paternity probabilities for each locus are calculated, from the built networks, by a software for inference on Bayesian Networks and these probabilities are combined into a single \"paternity index\".

A motivação para este trabalho originou-se da atual tendência do Gás Natural Liquefeito (GNL) se tornar uma importante opção para a diversificação da matriz energética brasileira. Atualmente, os terminais de gás natural liquefeito (GNL) são na maioria estruturas onshore, a construção destes terminais é custosa e muitos investimentos são necessários para atender as legislações ambientais e de segurança. Além disso, um acidente em uma destas instalações poderá produzir um grande impacto em áreas adjacentes. Sob esta perspectiva, surge uma nova proposta: uma unidade flutuante de armazenagem e regaseificação de gás natural liquefeito (FSRU - Floating Storage and Regasification Unit), o qual é uma unidade offshore e que pode trabalhar a quilômetros de distância da costa. O objetivo desta pesquisa é desenvolver uma metodologia de análise de Confiabilidade com o uso de Redes Bayesianas (RB) e aplicá-la na análise do sistema de Regaseificação do FSRU. O uso de RB, entre outras vantagens, permite a representação de incertezas no modelo e de dependências condicionais o que não é possível com as técnicas tradicionais, como por exemplo, as árvores de falhas e de eventos. Como resultado do trabalho, além da apresentação da metodologia a ser desenvolvida, serão identificados os pontos críticos do sistema contribuindo para o desenvolvimento de um plano de manutenção que assegure uma boa operabilidade do sistema com níveis razoáveis de dependabilidade. / The motivation for this research is the propensity of the Liquefied Natural Gas (LNG) becomes an important source of energy. Nowadays, LNG Import Terminals are mostly onshore; the construction of these terminals is costly and many adaptations are necessary to abide by environmental and safety laws. Moreover, an accident in one of these plants might produce considerable impact in neighboring areas. Under this perspective, a new option arises: a vessel known as FSRU (Floating Storage and Regasification Unit), which is an offshore unit, that can work miles away from de coast. The goal is to develop a Bayesian Network Reliability Modeling, which will show a preview of FSRUs Regasification System behavior. Using BN is possible to represent uncertain knowledge and local conditional dependencies. The results intend to clarify the critical equipment of the system and might improve the development of an effective maintenance plan, which can provide good operability with reasonable dependability levels.

Les réseaux de télécommunications deviennent de plus en plus complexes, notamment de par la multiplicité des technologies mises en œuvre, leur couverture géographique grandissante, la croissance du trafic en quantité et en variété, mais aussi de par l’évolution des services fournis par les opérateurs. Tout ceci contribue à rendre la gestion de ces réseaux de plus en plus lourde, complexe, génératrice d’erreurs et donc coûteuse pour les opérateurs. On place derrière le terme « réseaux autonome » l’ensemble des solutions visant à rendre la gestion de ce réseau plus autonome. L’objectif de cette thèse est de contribuer à la réalisation de certaines fonctions autonomiques dans les réseaux de télécommunications. Nous proposons une stratégie pour automatiser la gestion des pannes tout en couvrant les différents segments du réseau et les services de bout en bout déployés au-dessus. Il s’agit d’une approche basée modèle qui adresse les deux difficultés du diagnostic basé modèle à savoir : a) la façon d'obtenir un tel modèle, adapté à un réseau donné à un moment donné, en particulier si l'on souhaite capturer plusieurs couches réseau et segments et b) comment raisonner sur un modèle potentiellement énorme, si l'on veut gérer un réseau national par exemple. Pour répondre à la première difficulté, nous proposons un nouveau concept : l’auto-modélisation qui consiste d’abord à construire les différentes familles de modèles génériques, puis à identifier à la volée les instances de ces modèles qui sont déployées dans le réseau géré. La seconde difficulté est adressée grâce à un moteur d’auto-diagnostic actif, basé sur le formalisme des réseaux Bayésiens et qui consiste à raisonner sur un fragment du modèle du réseau qui est augmenté progressivement en utilisant la capacité d’auto-modélisation: des observations sont collectées et des tests réalisés jusqu’à ce que les fautes soient localisées avec une certitude suffisante. Cette approche de diagnostic actif a été expérimentée pour réaliser une gestion multi-couches et multi-segments des alarmes dans un réseau IMS. / While modern networks and services are continuously growing in scale, complexity and heterogeneity, the management of such systems is reaching the limits of human capabilities. Technically and economically, more automation of the classical management tasks is needed. This has triggered a significant research effort, gathered under the terms self-management and autonomic networking. The aim of this thesis is to contribute to the realization of some self-management properties in telecommunication networks. We propose an approach to automatize the management of faults, covering the different segments of a network, and the end-to-end services deployed over them. This is a model-based approach addressing the two weaknesses of model-based diagnosis namely: a) how to derive such a model, suited to a given network at a given time, in particular if one wishes to capture several network layers and segments and b) how to reason a potentially huge model, if one wishes to manage a nation-wide network for example. To address the first point, we propose a new concept called self-modeling that formulates off-line generic patterns of the model, and identifies on-line the instances of these patterns that are deployed in the managed network. The second point is addressed by an active self-diagnosis engine, based on a Bayesian network formalism, that consists in reasoning on a progressively growing fragment of the network model, relying on the self-modeling ability: more observations are collected and new tests are performed until the faults are localized with sufficient confidence. This active diagnosis approach has been experimented to perform cross-layer and cross-segment alarm management on an IMS network.