No description available.

Les Environnements Informatiques pour l'Apprentissage Humain ont été utilisés pour améliorer l'apprentissage humain. Ils visent à accroître la performance des élèves en fournissant un enseignement individualisé. Il a été reconnu que l'apprentissage individualisé est plus efficace que l'apprentissage classique. L'utilisation de modèles d'étudiants pour capturer les connaissances des élèves sous-tend l'apprentissage individualisé. Différents modèles d'étudiants ont été proposés. Toutefois, une partie des informations de diagnostic issues du comportement des élèves est généralement ignorée par ces modèles. En outre, pour individualiser les parcours d'apprentissage des élèves, les modèles d'étudiants devraient capturer les structures préalables de compétences. Toutefois, l'acquisition de structures de compétences nécessite beaucoup d'efforts d'ingénierie de la connaissance. Nous améliorons les modèles d'étudiants pour l'apprentissage individualisé selon deux aspects. D'une part, afin d'améliorer la capacité de diagnostic d'un modèle de l'élève, nous introduisons les motifs d'erreur d'étudiants. Pour traiter le bruit dans les données de performance des élèves, nous étendons un modèle probabiliste en y intégrant les réponses erronées. Les résultats montrent que la fonction de diagnostic permet d'améliorer la précision de la prédiction des modèles d'étudiant. D'autre part, nous cherchons à découvrir des structures de compétences préalables à partir des données de performance de l'élève. C'est une tâche difficile, car les connaissances des élèves constituent une variable latente. Nous proposons une méthode en deux phases. Notre procédé est validé en l'appliquant à des données. / Computer-based educational environments, like Intelligent Tutoring Systems (ITSs), have been used to enhance human learning. These environments aim at increasing student achievement by providing individualized instructions. It has been recognized that individualized learning is more effective than the conventional learning. Student models which are used to capture student knowledge underlie the individualized learning. In recent decades, various competing student models have been proposed. However, some diagnostic information in student behaviors is usually ignored by these models. Furthermore, to individualize learning paths, student models should capture prerequisite structures of fine-grained skills. However, acquiring skill structures requires much knowledge engineering effort. We improve student models for individualized learning with respect to the two aspects. On one hand, in order to improve the diagnostic ability of a student model, we introduce the diagnostic feature—student error patterns. To deal with the noise in student performance data, we extend a sound probabilistic model to incorporate erroneous responses. The results show that the diagnostic feature improves the prediction accuracy of student models. On the other hand, we target on discovering prerequisite structures of skills from student performance data. It is a challenging task, since student knowledge of a skill is a latent variable. We propose a two-phase method to discover skill structure from noisy observations. Our method is validated on simulated data and real data. In addition, we verify that prerequisite structures of skills can improve the accuracy of a student model.

In this paper, a study is conducted to investigate the use of Curriculum Learning as an approach to address accuracy issues in a neural network caused by training on a Long-Tailed dataset. The thesis problem is presented by a Swedish e-commerce company. Currently, they are using a neural network that has been modified by them using a CORAL framework. This adaptation means that instead of having a classic binary regression model, it is an ordinal regression model. The data used for training the model has a Long-Tail distribution, which leads to inaccuracies when predicting a price distribution for items that are part of the tail-end of the data. The current method applied to remedy this problem is Re-balancing in the form of down-sampling and up-sampling. A linear training scheme is introduced, increasing in increments of $10\%$ while applying Curriculum Learning. As a method for sorting the data in an appropriate way, inspiration is drawn from Knowledge Distillation, specifically the Teacher-Student model approach. The teacher models are trained as specialists on three different subsets, and furthermore, those models are used as a basis for sorting the data before training the student model. During the training of the student model, the Curriculum Learning approach is used. The results show that for Imbalance Ratio, Kullback-Liebler divergence, Class Balance, and the Gini Coefficient, the data is clearly less Long-Tailed after dividing the data into subsets. With the correct settings before training, there is also an improvement in the training speed of the student model compared to the base model. The accuracy for both the student model and the base model is comparable. There is a slight advantage for the base model when predicting items in the head part of the data, while the student model shows improvements for items that are between the head and the tail. / I denna uppsats genomförs en studie för att undersöka användningen av Curriculum Learning som en metod för att hantera noggrannhetsproblem i ett neuralt nätverk som är en konsekvens av träning på data som har en Long-Tail fördelning. Problemstälnningen som behandlas i uppsatsen är tillhandagiven av ett svensk e-handelsföretag. För närvarande använder de ett neuralt nätverk som har modifierats med hjälp av ett CORAL-ramverk. Denna anpassning innebär att det istället för att ha en klassisk binär regressionsmodell har en ordinal regressionsmodell. Datan som används för att träna modellen har en Long-Tail fördelning, vilket leder till problem vid prediktering av prisfördelning för diverse föremål som tillhör datans svans. Den nuvarande metod som används för att åtgärda detta problem är en Re-balancing i form av down-sampling och up-sampling. Ett linjärt träningschema introduceras, som ökar i steg om $10\%$ medan Curriculum Learning tillämpas. Metoden för att sortera datan på ett lämpligt sätt inspires av Knowledge-Distillation, mer specifikt lärar-elevmodell delen. Lärarmodellerna tränas som specialister på tre olika delmängder, och därefter används dessa modeller som grund för att sortera datan innan tränandet av elevmodellen. Under träningen av elevmodellen tillämpas Curriculum Learning. Resultaten visar att för Imbalance Ratio, Kullback-Libler-divergens, Class Balance och Gini-koefficienten är datat tydligt mindre Long-Tailed efter att datat delats in i delmängder. Med rätt inställningar innan tränandet finns även en förbättring i träningshastighet för elevmodellen jämfört med basmodellen. Noggrannheten för både elevmodellen och basmodellen är jämförbar. Det finns en liten fördel för basmodellen vid prediktering av föremål i huvuddelen av datan, medan elevmodellen visar förbättringar för föremål som ligger mellan huvuddelen och svansen.

A trace of lifelong-learning qualifications has become more mandatory at the European and even at world level. However, for higher education courses, the former could imply complex learning designs and abundance of data to monitor, analyze, and report. This work combine the ideas of personalized, competence-based, and social learning by providing course lifecycle support through competence-based design, outcome based assessment, social learning context analytics, and open student modeling visualizations. A series of studies using a virtual learning environment exploited the idea of the approach and revealed promising results. These results demonstrated the approach helped students and teachers to trace learning outcomes of the European Qualifications Framework (EQF) in higher education courses. Thus, this thesis extends the approach of higher education to a larger collection of learning objects for designing, assessing, and analyzing courses. Moreover, this approach verifies its capability of supporting social context visualization for online and blended personalized education. / Un rastreo del aprendizaje a lo largo de la vida es cada vez más obligatorio en Europa y el mundo. Sin embargo, en educación superior, esto implica diseños de cursos más complejos y abundancia de datos para rastrear, analizar e informar. Este trabajo combina aprendizaje personalizado, basado en competencias y social, apoyando diferentes instantes de los cursos universitarios a través de diseño basado en competencias, evaluación basada en resultados, analíticas del contexto social del aprendizaje y modelos abiertos del estudiante. Una serie de estudios exploró el enfoque revelando resultados prometedores. Se demostró el apoyo a estudiantes y maestros para monitorizar el aprendizaje según el EQF en sus cursos. Así, esta tesis extiende la educación superior a una colección más grande de objetos de aprendizaje para el diseño, la evaluación y el análisis de cualificaciones. Además, brinda visualización del contexto social para educación personalizada en ambientes mixtos y en línea.