Les réseaux métaboliques à l’échelle génomique décrivent les relations entre milliers de réactions et molécules biochimiques pour améliorer notre compréhension du métabolisme. Ils trouvent des applications dans les domaines chimiques, pharmaceutiques, et dans la biorestauration.La complexité de modèles métaboliques mets des obstacles á l’inférence des modèles, à la comparaison entre eux, ainsi que leur analyse, curation et amélioration par des experts humains. Parce que l’abondance des détailles dans les réseaux à grande échelle peut cacher des erreurs et des adaptations importantes de l’espèce qui est étudié, c’est important de trouver les correct niveaux d’abstraction qui sont confortables pour les experts humains : on doit mettre en évidence la structure essentiel du modèle ainsi que les divergences de celle-là (par exemple les chemins alternatives et les réactions manquantes), tout en masquant les détails non significatifs.Pour répondre a cette demande nous avons défini une généralisation des modèles métaboliques, fondée sur les connaissances, qui permet la création des vues abstraites de réseaux métaboliques. Nous avons développé une méthode théorétique qui regroupe les métabolites en classes d’équivalence et factorise les réactions reliant ces classes d’équivalence. Nous avons réalisé cette méthode comme une bibliothèque Python qui peut être téléchargée depuis metamogen.gforge.inria.fr.Pour valider l’intérêt de notre méthode, nous l’avons appliquée à 1 286 modèles métaboliques que nous avons extraits de la ressource Path2Model. Nous avons montré que notre méthode aide l’expert humain à relever de façon automatique les adaptations spécifiques de certains espèces et à comparer les modèles entre eux.Après en avoir discuté avec des utilisateurs, nous avons décidé de définir trois niveaux hiérarchiques de représentation de réseaux métaboliques : les compartiments, les modules et les réactions détaillées. Nous avons combiné notre méthode de généralisation et le paradigme des interfaces zoomables pour développer Mimoza, un système de navigation dans les réseaux métaboliques qui crée et visualise ces trois niveaux. Mimoza est accessible en ligne et pour le téléchargement depuis le site mimoza.bordeaux.inria.fr. / Genome-scale metabolic models describe the relationships between thousands of reactions and biochemical molecules, and are used to improve our understanding of organism’s metabolism. They found applications in pharmaceutical, chemical and bioremediation industries.The complexity of metabolic models hampers many tasks that are important during the process of model inference, such as model comparison, analysis, curation and refinement by human experts. The abundance of details in large-scale networks can mask errors and important organism-specific adaptations. It is therefore important to find the right levels of abstraction that are comfortable for human experts. These abstract levels should highlight the essential model structure and the divergences from it, such as alternative paths or missing reactions, while hiding inessential details.To address this issue, we defined a knowledge-based generalization that allows for production of higher-level abstract views of metabolic network models. We developed a theoretical method that groups similar metabolites and reactions based on the network structure and the knowledge extracted from metabolite ontologies, and then compresses the network based on this grouping. We implemented our method as a python library, that is available for download from metamogen.gforge.inria.fr.To validate our method we applied it to 1 286 metabolic models from the Path2Model project, and showed that it helps to detect organism-, and domain-specific adaptations, as well as to compare models.Based on discussions with users about their ways of navigation in metabolic networks, we defined a 3-level representation of metabolic networks: the full-model level, the generalized level, the compartment level. We combined our model generalization method with the zooming user interface (ZUI) paradigm and developed Mimoza, a user-centric tool for zoomable navigation and knowledgebased exploration of metabolic networks that produces this 3-level representation. Mimoza is available both as an on-line tool and for download atmimoza.bordeaux.inria.fr.

In aposematism a prey species use bright colours, often combined with a black contrasting pattern, to signal unprofitability as prey to potential predators. Although there are several different hypotheses about the presence of these internally contrasting patterns, there is little experimental evidence of any beneficial effects. In this thesis I have used bird predators and artificial prey signals to investigate if the contrasting internal patterns in warning displays may have evolved to increase signal efficacy, especially regarding the speed of avoidance learning. In paper I the relative importance of colour and pattern in avoidance learning was studied. The conclusion was that birds primarily attend to colour, not pattern, when learning the discrimination, which was further supported by the results in paper II-IV, all suggesting a secondary role of patterns. In paper II I show that predators may to some degree use patterns for discrimination, if they convey important information about prey quality. The predators showed a hierarchical way of learning warning colour components, where colour is learned to a higher degree than pattern. In paper III I investigate if internal contrasting patterns promote avoidance learning by increasing conspicuousness as prey-to-background contrast does. The study did not support this idea, as the presence of internal black patterns did not improve avoidance learning on a colour matching background. In paper IV, however, I show that the presence of many internal colour boundaries resulted in faster avoidance learning on a multi-coloured background, and predator generalization favoured more internal boundaries, while there was no effect of pattern regularity. From these studies I conclude that internal pattern contrasts may function to increase the efficacy of the warning colour, its salience, and as a means for aposematic prey to be discriminated from harmful mimics. However, the major finding is the importance of colour over pattern. / <strong><em></em></strong><em></em><em></em>At  the time of the doctoral defence the following papers were unpublished and had a status as follows: Paper 2: Manuscript; Paper 4: Manuscript<strong><em> </em></strong>

This thesis demonstrates how an interdisciplinary, organic music program built on the theory of Alfred North Whitehead and the practice of Carl Orff fosters freedom, imagination, creative expression, discipline, and wisdom so that children may learn about themselves, others, and the living world through an interdisciplinary approach to music. The thesis also addresses the failure of music educators to fully recognize and embrace the compatibility of Orff and Whiteheads work and the subsequent deficiency of school systems to provide aesthetic experiences through elementary music programs that develop wisdom.<p> The thesis underlines the importance of education that enables balanced growth in children; provides an historical perspective on the factors that contributed to the development of Orffs Schulwerk; emphasizes the importance of arts education in Whiteheads educational philosophy; and provides a constructive proposal for a music education program utilizing the philosophy and practice of Whitehead and Orff. The potentiality for music education and a process-based approach combining the work of Orff and Whitehead designed to nurture growth and wisdom and help children live life to the fullest, are underscored.

This study investigated age differences in generalization of causal value employing similarity as a cue to causality. Exemplars from six food categories (A+, B-, C+, D-. E+, F-) were presented to both young and older adults in two contiguous training phases. Training Phase 1 included exemplars from categories A+, B-, C+, D-. Training Phase 2 included exemplars from A+, B-, E+, F-. Foods in the “+” categories were paired with an outcome of sickness and foods in the “-” categories were not paired with sickness. Tests of causal judgment and exemplar recognition were conducted. For causal judgment, individual exemplars experienced during training and novel exemplars from all six categories were presented. For categories A+ and B-, the categories experienced in both training phases, young and older groups generalized the causal value to the category label and to all exemplars regardless of whether they were experienced in training or were novel. For categories experienced only once in training (C+, D-, E+, F-), both groups were better able to successfully judge causal value for experienced exemplars than novel exemplars. For young and older adults, experience made a difference in the ability to generalize causal value. Experienced and novel exemplars were also presented for recognition. Participants in both age groups showed a false memory effect for individual exemplars from the more experienced categories (A+, B-) suggesting that the process that allowed them to generalize causal value also interfered with their memory for individual exemplars. There was a difference between the younger and older groups for the categories that were only experienced once in training (C+, D-, E+, F-). In this case, younger participants showed better recognition than older adults for the individual exemplars. Older adults showed the same false memory effects for these categories as they showed for categories A+ and B-. These findings suggest that older adults generalize causal value as well as younger adults, but they are less able to distinguish individual exemplars. This discrepancy may be explained by differences in ability to use verbatim and gist. Older adults’ reduced verbatim processing leads to default gist encoding that enables them to focus on category level features but not process detailed exemplar identity (Brainerd & Reyna, 1990). Younger adults appear to have a flexibility that enables them to encode and retrieve both category-level gist and verbatim individualexemplar features when the task calls for it.

This thesis demonstrates how an interdisciplinary, organic music program built on the theory of Alfred North Whitehead and the practice of Carl Orff fosters freedom, imagination, creative expression, discipline, and wisdom so that children may learn about themselves, others, and the living world through an interdisciplinary approach to music. The thesis also addresses the failure of music educators to fully recognize and embrace the compatibility of Orff and Whiteheads work and the subsequent deficiency of school systems to provide aesthetic experiences through elementary music programs that develop wisdom.<p> The thesis underlines the importance of education that enables balanced growth in children; provides an historical perspective on the factors that contributed to the development of Orffs Schulwerk; emphasizes the importance of arts education in Whiteheads educational philosophy; and provides a constructive proposal for a music education program utilizing the philosophy and practice of Whitehead and Orff. The potentiality for music education and a process-based approach combining the work of Orff and Whitehead designed to nurture growth and wisdom and help children live life to the fullest, are underscored.

In pattern recognition, &ldquo / bias-variance&rdquo / trade-off is a challenging issue that the scientists has been working to get better generalization performances over the last decades. Among many learning methods, two-layered homogeneous stacked generalization has been reported to be successful in the literature, in different problem domains such as object recognition and image annotation. The aim of this work is two-folded. First, the problems of stacked generalization are attacked by a proposed novel architecture. Then, a set of success criteria for stacked generalization is studied. A serious drawback of stacked generalization architecture is the sensitivity to curse of dimensionality problem. In order to solve this problem, a new architecture named &ldquo / unanimous decision&rdquo / is designed. The performance of this architecture is shown to be comparably similar to two layered homogeneous stacked generalization architecture in low number of classes while it performs better than stacked generalization architecture in higher number of classes. Additionally, a new success criterion for two layered homogeneous stacked generalization architecture is proposed based on the individual properties of the used descriptors and it is verified in synthetic datasets.

Stacked Generalization (SG) is an ensemble learning technique, which aims to increase the performance of individual classifiers by combining them under a hierarchical architecture. This study consists of two major parts. In the first part, the performance of Stacked Generalization technique is analyzed with respect to the performance of the individual classifiers and the content of the training data. In the second part, based on the findings for a new class of algorithms, called Meta-Fuzzified Yield Value (Meta-FYV) is introduced. The first part introduces and verifies two hypotheses by a set of controlled experiments to assure the performance gain for SG. The learning mechanisms of SG to achieve high performance are explored and the relationship between the performance of the individual classifiers and that of SG is investigated. It is shown that if the samples in the training set are correctly classified by at least one base layer classifier, then, the generalization performance of the SG is increased, compared to the performance of the individual classifiers. In the second hypothesis, the effect of the spurious samples, which are not correctly labeled by any of the base layer classifiers, is investigated. In the second part of the thesis, six theorems are constructed based on the analysis of the feature spaces and the stacked generalization architecture. Based on the theorems and hypothesis, a new class of SG algorithms is proposed. The experiments are performed on both Corel data and synthetically generated data, using parallel programming techniques, on a high performance cluster.

In this thesis the action learning and generation problem on a humanoid robot is studied. Our aim is to realize action learning, generation and recognition in one system and our inspiration source is the mirror neuron hypothesis which suggests that action learning, generation and recognition share the same neural circuitry. Dynamic Movement Primitives, an efficient action learning and generation approach, are modified in order to fulfill this aim. The system we developed (1) can learn from multiple demonstrations, (2) can generalize to different conditions, (3) generates actions in a closed-loop and online fashion and (4) can be used for online action recognition. These claims are supported by experiments and the applicability of the developed system in real world is demonstrated through implementing it on a humanoid robot.

This study involved families of young children with autism spectrum disorders to examine the feasibility of implementing an adapted version of the school-based Prevent-Teach- Reinforce (PTR) model. This research included two families who developed and implemented the intervention for their children in collaboration with the researcher. The PTR manual was modified for use in a family context. The PTR intervention was tested using a multiple baseline design across routines. Procedural fidelity was assessed during training and coaching, as well as family implementation fidelity and social validity. To examine the potential efficacy of the adapted PTR intervention, the children’s target problem behavior and functionally equivalent alternative behavior were measured using video observation across experimental conditions including a generalization probe. Results indicated that the adapted PTR model is associated with reduction in child problem behavior and increases in alternative behavior. This study expanded the current research on the PTR model and extended its use to a novel setting and population so that a standardized model for positive behavior support implementation can be developed in the family context.

The purpose of this study was to compare the effects of two feedback procedures: written feedback and written feedback plus audio feedback on two teachers’ use of behavior support strategies and two children’s problem and alternative behaviors in two community preschool settings. A non-concurrent multiple-baseline design across teacher-child dyads with an A-B-C sequence was used to assess the effects of the feedback procedures on teacher and child target behaviors. A 10-second partial interval recording system was used to measure child target behaviors during 10-minute sessions and an event recording system to measure teacher use of strategies. The results indicated that the written feedback increased teachers’ use of support strategies and reduced children’s problem behaviors and increased alternative behaviors; however, the audio feedback with written feedback procedures further increased teachers’ use of strategies resulting in further improvement in children’s target behaviors. There was some evidence that teachers maintained their use of strategies without feedback procedures and generalized the use of strategies to non-targeted children.