Cette thèse a pour but d’améliorer l’automatisation dans l’ingénierie dirigée par les modèles (MDE pour Model Driven Engineering). MDE est un paradigme qui promet de réduire la complexité du logiciel par l’utilisation intensive de modèles et des transformations automatiques entre modèles (TM). D’une façon simplifiée, dans la vision du MDE, les spécialistes utilisent plusieurs modèles pour représenter un logiciel, et ils produisent le code source en transformant automatiquement ces modèles. Conséquemment, l’automatisation est un facteur clé et un principe fondateur de MDE. En plus des TM, d’autres activités ont besoin d’automatisation, e.g. la définition des langages de modélisation et la migration de logiciels. Dans ce contexte, la contribution principale de cette thèse est de proposer une approche générale pour améliorer l’automatisation du MDE. Notre approche est basée sur la recherche méta-heuristique guidée par les exemples. Nous appliquons cette approche sur deux problèmes importants de MDE, (1) la transformation des modèles et (2) la définition précise de langages de modélisation. Pour le premier problème, nous distinguons entre la transformation dans le contexte de la migration et les transformations générales entre modèles. Dans le cas de la migration, nous proposons une méthode de regroupement logiciel (Software Clustering) basée sur une méta-heuristique guidée par des exemples de regroupement. De la même façon, pour les transformations générales, nous apprenons des transformations entre modèles en utilisant un algorithme de programmation génétique qui s’inspire des exemples des transformations passées. Pour la définition précise de langages de modélisation, nous proposons une méthode basée sur une recherche méta-heuristique, qui dérive des règles de bonne formation pour les méta-modèles, avec l’objectif de bien discriminer entre modèles valides et invalides. Les études empiriques que nous avons menées, montrent que les approches proposées obtiennent des bons résultats tant quantitatifs que qualitatifs. Ceux-ci nous permettent de conclure que l’amélioration de l’automatisation du MDE en utilisant des méthodes de recherche méta-heuristique et des exemples peut contribuer à l’adoption plus large de MDE dans l’industrie à là venir. / This thesis aims to improve automation in Model Driven Engineering (MDE). MDE is a paradigm that promises to reduce software complexity by the mean of the intensive use of models and automatic model transformation (MT). Roughly speaking, in MDE vision, stakeholders use several models to represent the software, and produce source code by automatically transforming these models. Consequently, automation is a key factor and founding principle of MDE. In addition to MT, other MDE activities require automation, e.g. modeling language definition and software migration. In this context, the main contribution of this thesis is proposing a general approach for improving automation in MDE. Our approach is based on meta-heuristic search guided by examples. We apply our approach to two important MDE problems, (1) model transformation and (2) precise modeling languages. For transformations, we distinguish between transformations in the context of migration and general model transformations. In the case of migration, we propose a software clustering method based on a search algorithm guided by cluster examples. Similarly, for general transformations, we learn model transformations by a genetic programming algorithm taking inspiration from examples of past transformations. For the problem of precise metamodeling, we propose a meta-heuristic search method to derive well-formedness rules for metamodels with the objective of discriminating examples of valid and invalid models. Our empirical evaluation shows that the proposed approaches exhibit good results. These allow us to conclude that improving automation in MDE using meta-heuristic search and examples can contribute to a wider adoption of MDE in industry in the coming years.

En la actualidad, el uso de las tecnologías ha sido primordial para el avance de las sociedades, estas han permitido que personas sin conocimientos informáticos o usuarios llamados “no expertos” se interesen en su uso, razón por la cual los investigadores científicos se han visto en la necesidad de producir estudios que permitan la adaptación de sistemas, a la problemática existente dentro del ámbito informático. Una necesidad recurrente de todo usuario de un sistema es la gestión de la información, la cual se puede administrar por medio de una base de datos y lenguaje específico, como lo es el SQL (Structured Query Language), pero esto obliga al usuario sin conocimientos a acudir a un especialista para su diseño y construcción, lo cual se ve reflejado en costos y métodos complejos, entonces se plantea una pregunta ¿qué hacer cuando los proyectos son pequeñas y los recursos y procesos son limitados? Teniendo como base la investigación realizada por la universidad de Washington[39], donde sintetizan sentencias SQL a partir de ejemplos de entrada y salida, se pretende con esta memoria automatizar el proceso y aplicar una técnica diferente de aprendizaje, para lo cual utiliza una aproximación evolucionista, donde la aplicación de un algoritmo genético adaptado origina sentencias SQL válidas que responden a las condiciones establecidas por los ejemplos de entrada y salida dados por el usuario. Se obtuvo como resultado de la aproximación, una herramienta denominada EvoSQL que fue validada en este estudio. Sobre los 28 ejercicios empleados por la investigación [39], 23 de los cuales se obtuvieron resultados perfectos y 5 ejercicios sin éxito, esto representa un 82.1% de efectividad. Esta efectividad es superior en un 10.7% al establecido por la herramienta desarrollada en [39] SQLSynthesizer y 75% más alto que la herramienta siguiente más próxima Query by Output QBO[31]. El promedio obtenido en la ejecución de cada ejercicio fue de 3 minutos y 11 segundos, este tiempo es superior al establecido por SQLSynthesizer; sin embargo, en la medida un algoritmo genético supone la existencia de fases que amplían los rangos de tiempos, por lo cual el tiempo obtenido es aceptable con relación a las aplicaciones de este tipo. En conclusión y según lo anteriormente expuesto, se obtuvo una herramienta automática con una aproximación evolucionista, con buenos resultados y un proceso simple para el usuario “no experto”. / Actuellement l'usage des technologies est primordial pour l'avance de la société, celles-ci ont permis que des personnes sans connaissances informatiques ou des utilisateurs appelés "non expert" s'intéressent à son usage. C'est la raison pour laquelle les enquêteurs scientifiques se sont vus dans la nécessité de produire les études qui permettent l'adaptation des systèmes à la problématique existante à l'intérieur du domaine informatique. Une nécessité récurrente pour tout utilisateur d'un système est la gestion de l'information, que l’on peut administrer au moyen d'une base de données et de langage spécifique pour celles-ci comme est le SQL (Structured Query Language), mais qui oblige à l'utilisateur à chercher un spécialiste pour sa conception et sa construction, et qui représente des prix et des méthodes complexes. Une question se pose alors, quoi faire quand les projets sont petites et les ressources et les processus limités ? Ayant pour base la recherche de l'université de Washington [39], ce mémoire automatise le processus et applique une différente technique d'apprentissage qui utilise une approche évolutionniste, où l'application d'un algorithme génétique adapté génère des requêtes SQL valides répondant aux conditions établies par les exemples d'entrée et de sortie donnés par l'utilisateur. On a obtenu comme résultat de l’approche un outil dénommé EvoSQL qui a été validé dans cette étude. Sur les 28 exercices employés par la recherche [39], 23 exercices ont été obtenus avec des résultats parfaits et 5 exercices sans succès, ce qui représente 82.1 % d'effectivité. Cette effectivité est supérieure de 10.7 % à celle établie par l'outil développé dans [32] SQLSynthesizer et 75% plus haute que l'outil suivant le plus proche Query by Output QBO [31]. La moyenne obtenue dans l'exécution de chaque exercice a été de 3 min et 11sec, ce qui est supérieur au temps établi par SQlSynthesizer, cependant dans la mesure où un algorithme génétique suppose que l'existence de phases augmente les rangs des temps, le temps obtenu est acceptable par rapport aux applications de ce type. Dans une conclusion et selon ce qui a été antérieurement exposé nous avons obtenu un outil automatique, avec une approche évolutionniste, avec de bons résultats et un processus simple pour l'utilisateur « non expert ». / At present the use of the technologies is basic for the advance of the society; these have allowed that persons without knowledge or so called "non expert" users are interested in this use, is for it that the researchers have seen the need to produce studies that allow the adjustment of the systems the existing at the problematic inside the area of the technology. A need of every user of a system is the management of the information, which can be manage by a database and specific language for these as the SQL (Structured Query Language), which forces the user to come to a specialist for the design and construction of this one, which represents costs and complex methods, but what to do when they are small investigations where the resources and processes are limited? Taking as a base the research of the university of Washington [32], this report automates the process and applies a different learning technique, for which uses an evolutionary approach, where the application of a genetic adapted algorithm generates query SQL valid that answer to the conditions established by the given examples of entry and exit given by the user. There was obtained as a result of the approach a tool named EvoSQL that was validated in the same 28 exercises used by the investigation [32], of which 23 exercises were obtained by ideal results and 5 not successful exercises, which represents 82.1 % of efficiency, superior in 10.7 % to the established one for the tool developed in [32] SQLSynthesizer and 75% higher than the following near tool Query by Output QBO [26]. The average obtained in the execution of every exercise was of 3 min and 11seg that is superior to the time established by SQlSynthesizer, Nevertheless, being a genetic algorithm where the steps existence makes that the ranges of times are extended, the obtained one is acceptable with relation to the applications of this type. In conclusion et according to previously exposed, we have obtained an automatic tool, with an evolutionary approach, with good results and a simple process for the « not expert » user.

No description available.

La classification de séries temporelles a suscité beaucoup d’intérêt au cours des dernières années en raison de ces nombreuses applications. Nous commençons par proposer la méthode Dense Bag-of-Temporal-SIFT-Words (D-BoTSW) qui utilise des descripteurs locaux basés sur la méthode SIFT, adaptés pour les données en une dimension et extraits à intervalles réguliers. Des expériences approfondies montrent que notre méthode D-BoTSW surpassent de façon significative presque tous les classificateurs de référence comparés. Ensuite, nous proposons un nouvel algorithmebasé sur l’algorithme Learning Time Series Shapelets (LTS) que nous appelons Adversarially- Built Shapelets (ABS). Cette méthode est basée sur l’introduction d’exemples adversaires dans le processus d’apprentissage de LTS et elle permet de générer des shapelets plus robustes. Des expériences montrent une amélioration significative de la performance entre l’algorithme de base et notre proposition. En raison du manque de jeux de données labelisés, formatés et disponibles enligne, nous utilisons deux jeux de données appelés TiSeLaC et Brazilian-Amazon. / Time Series Classification (TSC) has received an important amount of interest over the past years due to many real-life applications. In this PhD, we create new algorithms for TSC, with a particular emphasis on Remote Sensing (RS) time series data. We first propose the Dense Bag-of-Temporal-SIFT-Words (D-BoTSW) method that uses dense local features based on SIFT features for 1D data. Extensive experiments exhibit that D-BoTSW significantly outperforms nearly all compared standalone baseline classifiers. Then, we propose an enhancement of the Learning Time Series Shapelets (LTS) algorithm called Adversarially-Built Shapelets (ABS) based on the introduction of adversarial time series during the learning process. Adversarial time series provide an additional regularization benefit for the shapelets and experiments show a performance improvementbetween the baseline and our proposed framework. Due to the lack of available RS time series datasets,we also present and experiment on two remote sensing time series datasets called TiSeLaCand Brazilian-Amazon

近年來，信用卡市場快速地成長，發卡銀行亦大量地發卡，然而目前國內 發卡銀行在整個信用卡信用風險管理上，大都採行人類專家經驗判斷的方 式進行。發卡銀行隨著持卡人數快速地增加，其信用資料亦呈等比例急速 上升，若仍採用人工處理方式，除了會大幅增加工作負荷外，其授信品質 也不易控制。因此，本研究擬引進資訊技術來解決大量信用卡信用資料之 信用管理問題。 首先，我們探討信用卡信用管理業務，並根據其作業 流程來建構一信用卡信用管理自動化的架構，此架構包括徵信驗證系統、 審核系統、預警系統、高風險客戶管理系統、及催收系統等五個系統，其 目的在於輔助授信管理之業務、減少授管人員的工作負荷、以有效控制授 信品質、及降低授信的風險。 其次，本研究針對上述信用卡信用管理 自動化中的預警系統，利用範例學習法來建立信用卡信用風險預警範例學 習系統，且實際以一家發卡銀行的信用資料來建立並驗證四個預警模式， 期能事先讓系統自動查核信用不良之客戶。此四類預警模式為: (一)提前 預警模式(二)群體決策預警模式(三)追蹤管理預警模式(四)例外管理預警 模式 最後，我們亦提出一些未來研究之課題，期能進一步發展本研究 之信用卡信用管理自動化系統及預警模式，以推廣應用至各發卡機構。

隨著國人持信用卡消費購物方式的蔚為風氣，致使發卡機構每日所需處理 的申請案件激增；同時，由於信用卡業務的日趨多元化，更增添了審核的 複雜度。傳統以人為判斷為主的審核方法，在有限的人力之下，勢將難以 因應如此龐大的審核需求，而在時間緊迫、經驗累積不足的情形下，難免 會危及授信品質，而增加了此項授信業務的風險。有鑑於此，本研究希望 能藉由範例學習法建立一信用卡信用風險審核模式，期能有效輔助信用卡 發卡審核作業，以降低授信風險，並提昇發卡機構的經營績效。本研究以 某發卡機構為研究對象。抽取個案機構81全年度，審核通過的資料作為研 究樣本。其中以截至82年度3月止，被強制停卡者之不良卡戶，計2,788筆 ；而仍繼續流通的正常卡戶，計有97,001筆，總計99,789筆，作為系統學 習及測試所需之資料。本研究首先針對信用卡審核問題的特性，探討範例 學習法的處理策略，我們將計質線索以相對風險的觀念轉換為順序尺度， 並使所建構的二元樹之葉節點（判斷法則）精簡了二分之一左右，和原分 類樹預測能力並無顯著差異。其次，我們進一步運用修剪策略，可將原判 斷法則數由230條減至26條，大幅的提升了執行效率；修剪策略的運用， 雖然降低了區別率，但卻將預測能力（命中率）由67.1%提升至72.58%。 亦即研究結果顯示，避免分類過細，有助於系統預測能力的提升。本研究 範例學習審核模式之預測能力達72.58%，較 Logistic Re- gression 審 核模式高出約6.49%。在重要性線索的選取上，二者具有相當的一致性； 研究結果顯示，原持有一般卡張數、金卡張數、教育程度、公司等級、職 級等為區別力較佳的信用要素。其中區別能力最強的因素為原一般卡持有 張數，張數愈多，信用風險愈高，而此因素為原審核模式所疏漏者，值得 授信人員警惕。此外，我們再將成本效益因素加入分類樹判斷法則，透過 此方式可調整授信的門檻，以增加發卡機構所能獲取的利潤。進一步考量 申請者的信用風險與所得，建立一信用額度核定的方式。研究結果顯示， 以此一模式授與信用額度較原機構之授與方式,高出約一仟二佰萬元淨收 益。此乃由於本模式能有效辨識出不良客戶（命中率為 80.58%），因而 大幅地降低了呆帳的損失。最後，我們綜合本研究的心得，提出一些未來 研究課題，期能使最適分類樹的產生更具效率，並且擴大研究的範圍，希 望能將信用卡範例學習系統推展至各發卡機構，並應用於信用管理的各層 面，以有效提昇信用卡經營效益。

This thesis examines the concept of image self-similarity and provides solutions to various associated inverse problems such as resolution enhancement and missing fractal codes. In general, many real-world inverse problems are ill-posed, mainly because of the lack of existence of a unique solution. The procedure of providing acceptable unique solutions to such problems is known as regularization. The concept of image prior, which has been of crucial importance in image modelling and processing, has also been important in solving inverse problems since it algebraically translates to the regularization procedure. Indeed, much recent progress in imaging has been due to advances in the formulation and practice of regularization. This, coupled with progress in optimization and numerical analysis, has yielded much improvement in computational methods of solving inverse imaging problems. Historically, the idea of self-similarity was important in the development of fractal image coding. Here we show that the self-similarity properties of natural images may be used to construct image priors for the purpose of addressing certain inverse problems. Indeed, new trends in the area of non-local image processing have provided a rejuvenated appreciation of image self-similarity and opportunities to explore novel self-similarity-based priors. We first revisit the concept of fractal-based methods and address some open theoretical problems in the area. This includes formulating a necessary and sufficient condition for the contractivity of the block fractal transform operator. We shall also provide some more generalized formulations of fractal-based self-similarity constraints of an image. These formulations can be developed algebraically and also in terms of the set-based method of Projection Onto Convex Sets (POCS). We then revisit the traditional inverse problems of single frame image zooming and multi-frame resolution enhancement, also known as super-resolution. Some ideas will be borrowed from newly developed non-local denoising algorithms in order to formulate self-similarity priors. Understanding the role of scale and choice of examples/samples is also important in these proposed models. For this purpose, we perform an extensive series of numerical experiments and analyze the results. These ideas naturally lead to the method of self-examples, which relies on the regularity properties of natural images at different scales, as a means of solving the single-frame image zooming problem. Furthermore, we propose and investigate a multi-frame super-resolution counterpart which does not require explicit motion estimation among video sequences.

This thesis examines the concept of image self-similarity and provides solutions to various associated inverse problems such as resolution enhancement and missing fractal codes. In general, many real-world inverse problems are ill-posed, mainly because of the lack of existence of a unique solution. The procedure of providing acceptable unique solutions to such problems is known as regularization. The concept of image prior, which has been of crucial importance in image modelling and processing, has also been important in solving inverse problems since it algebraically translates to the regularization procedure. Indeed, much recent progress in imaging has been due to advances in the formulation and practice of regularization. This, coupled with progress in optimization and numerical analysis, has yielded much improvement in computational methods of solving inverse imaging problems. Historically, the idea of self-similarity was important in the development of fractal image coding. Here we show that the self-similarity properties of natural images may be used to construct image priors for the purpose of addressing certain inverse problems. Indeed, new trends in the area of non-local image processing have provided a rejuvenated appreciation of image self-similarity and opportunities to explore novel self-similarity-based priors. We first revisit the concept of fractal-based methods and address some open theoretical problems in the area. This includes formulating a necessary and sufficient condition for the contractivity of the block fractal transform operator. We shall also provide some more generalized formulations of fractal-based self-similarity constraints of an image. These formulations can be developed algebraically and also in terms of the set-based method of Projection Onto Convex Sets (POCS). We then revisit the traditional inverse problems of single frame image zooming and multi-frame resolution enhancement, also known as super-resolution. Some ideas will be borrowed from newly developed non-local denoising algorithms in order to formulate self-similarity priors. Understanding the role of scale and choice of examples/samples is also important in these proposed models. For this purpose, we perform an extensive series of numerical experiments and analyze the results. These ideas naturally lead to the method of self-examples, which relies on the regularity properties of natural images at different scales, as a means of solving the single-frame image zooming problem. Furthermore, we propose and investigate a multi-frame super-resolution counterpart which does not require explicit motion estimation among video sequences.

A key challenge of the learning sciences is moving research results into practice. Educators on the front lines perceive little value in the outputs of education research and demand more “usable knowledge”. This work explores the potential instead of usable artifacts to translate knowledge into practice, adding scientists as stakeholders in an interaction design process. The contributions are two effective systems, the scientific and contextual principles in their design, and a research model for scientific research through interaction design. College student study practices are the domain chosen for the development of these methods. Iterative ethnographic fieldwork identified two systems that would be likely to advance both learning in practice and knowledge for applying the employed theories in general. Nudge was designed to improve students’ study time management by regularly emailing students with explicit recommended study activities. It reconceptualizes the syllabus into an interactive guide that fits into modern students' attention streams. Examplify was designed to improve how students learn from worked example problems by modularizing them into steps and scaffolding their metacognitive behaviors though problem-solving and self-explanation prompts. It combines these techniques in a way that is exceedingly easy to author, using existing answer keys and students' self-evaluations. Nudge and Examplify were evaluated experimentally over a full semester of a lecture-based introductory chemistry course. Nudge messages increased students’ sense of achievement and interacted with students’ existing time management skills to improve exam grades for poorer students. Among students who could choose whether to receive them, 80% did. Students with access to Examplify had higher exam scores (d=0.26), especially on delayed measures of learning (d=0.40). A key design decision in Examplify was not clearly resolvable by existing theory and so was tested experimentally by comparing two variants, one without prompts to solve the steps. The variant without problem solving was less effective (d=0.77) and less used, while usage rates of the variant with problem solving increased over time. These results support the use of the design methods employed and provide specific empirical recommendations for future designs of these and similar systems for implementing theory in practice.

This thesis deals with the solution of differential equations of the first degree. The work is intended to serve as a textbook (a collection of exercises) for students of teaching mathematics at lower secondary schools. Each chapter contains a summary of basic concepts, solved task models of the related topic, sorted by difficulty, and finally tasks assigned for independent practicing. This thesis aims to present basic knowledge about ways of solving differential equations of the first degree, including practical skills for their solution.