L'apprentissage actif apparaît comme un problème important dans différents contextes de l'apprentissage supervisé pour lesquels obtenir des données est une tâche aisée mais les étiqueter est coûteux. En règle générale, c’est une stratégie de requête, une heuristique gloutonne basée sur un critère de sélection qui recherche les données non étiquetées potentiellement les plus intéressantes pour former ainsi un ensemble d'apprentissage. Une stratégie de requête est donc une procédure d'échantillonnage biaisée puisqu'elle favorise systématiquement certaines observations s'écartant ainsi des modèles d'échantillonnages indépendants et identiquement distribués. L'hypothèse principale de cette thèse s'inscrit dans la réduction du biais introduit par le critère de sélection. La proposition générale consiste à réduire le biais en sélectionnant le sous-ensemble minimal d'apprentissage pour lequel l'estimation de la loi de probabilité est aussi proche que possible de la loi sous-jacente prenant en compte l’intégralité des observations. Pour ce faire, une nouvelle stratégie générale de requête pour l'apprentissage actif a été mise au point utilisant la théorie de l'Information. Les performances de la stratégie de requête proposée ont été évaluées sur des données réelles et simulées. Les résultats obtenus confirment l'hypothèse sur le biais et montrent que l'approche envisagée améliore l'état de l'art sur différents jeux de données. / Active Learning arises as an important issue in several supervised learning scenarios where obtaining data is cheap, but labeling is costly. In general, this consists in a query strategy, a greedy heuristic based on some selection criterion, which searches for the potentially most informative observations to be labeled in order to form a training set. A query strategy is therefore a biased sampling procedure since it systematically favors some observations by generating biased training sets, instead of making independent and identically distributed draws. The main hypothesis of this thesis lies in the reduction of the bias inherited from the selection criterion. The general proposal consists in reducing the bias by selecting the minimal training set from which the estimated probability distribution is as close as possible to the underlying distribution of overall observations. For that, a novel general active learning query strategy has been developed using an Information-Theoretic framework. Several experiments have been performed in order to evaluate the performance of the proposed strategy. The obtained results confirm the hypothesis about the bias, showing that the proposal outperforms the baselines in different datasets.

Increasingly, multimedia technology is permeating the educational arena worldwide, and many colleges and universities are moving towards the use of digital technology to enhance the teaching and learning process of both the students and educational practitioners (Kachian & Wieser, 1999:[online]; Mat, 2000:[online]). South Africa is a developing country that is undergoing radical social, political, economic and cultural changes and advances in computer technology have also dramatically changed the learning and teaching process and provided new learning opportunities and access to educational resources beyond those traditionally available. This research study describes a design experiment in which a multimedia learning environment (MMLE) was crafted for nursing students in neonatology at a university in Johannesburg, South Africa. At the outset, the integrated design principles derived from the constructivist perspectives on learning, multimedia learning design principles generated from Instructional Design Theory and the learning styles according to Kolb’s Learning Style Inventory, are established as a theoretical point of departure. This theoretical position led to the generation of a design framework that exploits the congruencies between constructivist perspectives on learning, the individual attributes of learners as defined by learning styles and multimedia design principles. The design experiment is conducted in five phases: the establishment of the design framework after an extensive literature review, the development of the MMLE, a pilot study, the final implementation and data analysis. Both quantitative and qualitative data are collected. The preliminary results of the study show that the students had an overwhelmingly positive experience of the MMLE, and that their preferred learning style had some influence on their experience. Little evidence has been found in the literature about the interaction between learning styles and constructivist learning principles for the design of multimedia learning and it is in this area that the study makes a contribution. The study also makes a contribution to the field of Nursing Science education, as it designs and develops multimedia learning materials, and assesses the value of those learning materials for learners which may be adopted in similar contexts within the broader South African context. / Prof. D. Van Der Westhuizen

Hierarchical task decompositions play an essential role in the design of complex simulation and decision systems, such as the ones that arise in video games. Game designers find it very natural to adopt a divide-and-conquer philosophy of specifying hierarchical policies, where decision modules can be constructed somewhat independently. The process of choosing the parameters of these modules manually is typically lengthy and tedious. The hierarchical reinforcement learning (HRL) field has produced elegant ways of decomposing policies and value functions using semi-Markov decision processes. However, there is still a lack of demonstrations in larger nonlinear systems with discrete and continuous variables. To narrow this gap between industrial practices and academic ideas, we address the problem of designing efficient algorithms to facilitate the deployment of HRL ideas in more realistic settings. In particular, we propose Bayesian active learning methods to learn the relevant aspects of either policies or value functions by focusing on the most relevant parts of the parameter and state spaces respectively. To demonstrate the scalability of our solution, we have applied it to The Open Racing Car Simulator (TORCS), a 3D game engine that implements complex vehicle dynamics. The environment is a large topological map roughly based on downtown Vancouver, British Columbia. Higher level abstract tasks are also learned in this process using a model-based extension of the MAXQ algorithm. Our solution demonstrates how HRL can be scaled to large applications with complex, discrete and continuous non-linear dynamics. / Science, Faculty of / Computer Science, Department of / Graduate

Data Stream mining is an important emerging topic in the data mining and machine learning domain. In a Data Stream setting, the data arrive continuously and often at a fast pace. Examples include credit cards transaction records, surveillances video streams, network event logs, and telecommunication records. Such types of data bring new challenges to the data mining research community. Specifically, a number of researchers have developed techniques in order to build accurate classification models against such Data Streams. Ensemble Learning, where a number of so-called base classifiers are combined in order to build a model, has shown some promise. However, a number of challenges remain. Often, the class labels of the arriving data are incorrect or missing. Furthermore, Data Stream algorithms may benefit from an online learning paradigm, where a small amount of newly arriving data is used to learn incrementally. To this end, the use of Active Learning, where the user is in the loop, has been proposed as a way to extend Ensemble Learning. Here, the hypothesis is that Active Learning would increase the performance, in terms of accuracy, ensemble size, and the time it takes to build the model. This thesis tests the validity of this hypothesis. Namely, we explore whether augmenting Ensemble Learning with an Active Learning component benefits the Data Stream Learning process. Our analysis indicates that this hypothesis does not necessarily hold for the datasets under consideration. That is, the accuracies of Active Ensemble Learning are not statistically significantly higher than when using normal Ensemble Learning. Rather, Active Learning may even cause an increase in error rate. Further, Active Ensemble Learning actually results in an increase in the time taken to build the model. However, our results indicate that Active Ensemble Learning builds accurate models against much smaller ensemble sizes, when compared to the traditional Ensemble Learning algorithms. Further, the models we build are constructed against small and incrementally growing training sets, which may be very beneficial in a real time Data Stream setting.

This thesis is a record of two research strands that have been intertwined during the development over a four year period of a classroom curriculum for an elective Geology 12 course in a new school. It discusses traditional belief systems identified as common to the practice of senior science and how one teacher wanted to challenge those beliefs to produce a working curriculum that would focus on long term learning within the framework of an externally prescribed curriculum and a provincially mandated external final exam that counted for 40% of the students mark. The teacher, working on her own in a portable for the first two years was in the unenviable position of being supplied with textbooks with a foreign focus and with supplies that as the title suggests were of little use over the long term. By Christmas of the first year a number of major problems had been identified, these problems falling into two major categories - developing strategies for long term learning that, within the operational constraints of grade 12, would enable the students to take far more responsibility for their own learning, and second, developing a science research programme for acquiring the resources, principally through field work, that were identified as being necessary for the programme. The major concerns within these two problem areas were identified and a four year timeline was developed for implementation. On the pedagogical side, after examining some of the literature on learning, particularly that around the area of cooperative learning that has had a substantial focus in recent years in a number of local school districts, reflecting on what worked for me in terms of my practice over 27 years of science teaching, I chose to focus on the Project for Enhancing Effective Learning (PEEL), out of Monash University, Australia as my working framework for learning. The process of developing this classroom curriculum was framed as a qualitative individual action research project over time as, within my professional life, there were no other teachers involved with the geology programme within the school, and at the same time being in a portable isolated me from my peers-l had no choice but to be self contained and self reliant. The pedagogical side of the process saw the evolvement of a programme that differed significantly in many ways from traditional senior science teaching. This is not to say that many teachers are not already reflecting on and trying to improve practice but for most of them it is through quiet reflection, discourse and evolution much as it had been for me until this time. For me this was the first time in my career that I was able to develop a programme from the very beginning. The thesis details the development of a multi-level learning strategy with an underlying theme being the development of more metacognitive students. The programme entails the identification of prior learning, reflective and collaborative practice, multiple processings of content and skills, peer assessment, and semi-formal reflective assessment. For many students, particularly during my first two years, most of these strategies were completely foreign to their cultural expectations of the teacher's role as dispenser of information to be regurgitated back through formal assessment. During the last two years these challenges to student thinking have been far less dramatic as I am now a known quantity in the school and the students taking my course expect to be working at becoming more independent long term learners. The programme is also built on the premise that for geology, relevant hands-on activities are an integral part of the learning process, and this other research strand is also explored and described. This is the story of the two research strands by which a semi-independent multi-level learning environment has been developed and implemented with a high degree of hands-on activities. Although a formal assessment of the programme is almost impossible to do within the constraints of my working environment, the personal feedback that I receive from the students, parents and colleagues indicates that it has been successful. / Education, Faculty of / Curriculum and Pedagogy (EDCP), Department of / Graduate

In South Africa, the Department of Education (DOE) via its South African Qualifications Authority (SAQA) mandates lecturers particularly at higher education level to deliver students that should be able to think critically and solve problems by the end of their undergraduate journey at any Higher Education Institution (HEI), whether public or private. HEIs have each taken their own approach on how to develop these competencies in their undergraduate students. This qualitative inductive case study focuses on understanding how eleven lecturers teaching at a private HEI in Midrand South Africa facilitate Active Learning in their classes, how they measure the success of Active Learning strategies and the support they have available to them by using semi-structured interviews and class observation data. Some of the findings highlight that these lecturers know exactly what Active Learning is even though most have never been officially trained. Six groups of different Active Learning strategies were identified including different questioning techniques, engagement via reading, engagement via writing, hands-on activities, use of technology and interaction with peers. Even though lecturers believed in Active Learning, evidence substantiating the effectiveness of their teaching methodology was mostly subjective. It was also found that lecturers had more support requirements than current support available and that the majority of current support was in the form of the immediate lecturer community. / Thesis (PhD)--University of Pretoria, 2017. / Science, Mathematics and Technology Education / PhD / Unrestricted

Under what conditions and in what situations is the telling of personal history and other stories an effective teaching tool?Storytelling has been used by many of the greatest teachers throughout history. Plato, Jesus and Gandi, used stories, parables and personal histories to educate students (Zabel 1991). In fact, storytelling has been referred to as the foundation of the teaching profession (Abrahamson 1998). In recent years, the use of storytelling has received attention from academic researchers and has been studied in several academic disciplines. It has been suggested that the use of storytelling in higher education settings increases student performance and recollection (Bryant & Harris 2011). However, few students have considered the potential for the use of storytelling in the accounting classroom.This archival research seeks to describe the value of storytelling as a pedagogical tool across academic disciplines, review the literature regarding the use of storytelling in other academic disciplines in higher education, synthesize the findings of existing research and describe the uses, benefits and difficulties with using storytelling in various accountancy classes across the curriculum, and suggest possible uses for storytelling in accountancy classes.

No description available.

No description available.

We study online active learning under the Bernstein condition for bounded general losses and offer a solution for online variance estimation. Our suggested algorithm is based on IWAL (Importance Weighted Active Learning) which utilizes the online variance estimation technique to shrink the hypothesis set. For our algorithm, we provide a fallback guarantee and prove that in the case that R(f*) is small, it will converge faster than passive learning, where R(f*) is the risk of the best hypothesis in the hypothesis class. Finally, in the special case of zero-one loss exponential improvement is achieved in label complexity over passive learning. / Graduate