Padrões de interacção num ambiente de aprendizagem colaborativa a distância : um estudo de caso com conteúdos estruturados segundo a teoria da flexibilidade cognitiva

A Globalização a que assistimos actualmente traz consigo exigências que a Sociedade deve responder de forma efectiva e adequada. O eLearning constitui, assim, uma realidade capaz de congregar esforços no sentido de permitir a construção de comunidades empenhadas em adquirir as competências necessárias para enfrentar os desafios propostos pela Globalização. É nesta perspectiva que apresentamos este estudo que procura, na sua essência, compreender o processo de interacção num ambiente de aprendizagem colaborativo a distância entre alunos de Línguas Clássicas. Cientes da importância de promover uma aprendizagem com significado para os alunos, foram desenvolvidos conteúdos que representaram o conhecimento segundo os pressupostos pedo-didácticos da Teoria da Flexibilidade Cognitiva de forma a serem trabalhados colaborativamente pelos participantes no fórum de discussão online Scaena. O trabalho desenvolvido pelos alunos decorreu ao longo de três sessões e foi integrado na disciplina de Tecnologia Educativa constante do programa curricular do 2º Semestre do 4º Ano da Licenciatura de Português, Latim e Grego, no ano lectivo de 2004-05. De índole qualitativa, a investigação efectuada privilegiou a análise de conteúdo a fim de proceder ao tratamento de dados. Para o efeito foi, ainda, utilizado o software de análise NUD*IST. Os resultados relevaram a ocorrência de padrões de interacção em todas as dimensões de análise, assim como tornaram evidente o processo de construção de conhecimento flexível numa plataforma de ensino online. Por último, os resultados apurados confirmam as mais-valias da utilização das Tecnologias da Informação e Comunicação para os Estudos Clássicos em contexto educativo. São, ainda, apresentadas sugestões para futuros estudos. ABSTRACT: The Globalisation we witness nowadays brings with it demands to which Society has to answer effectivelly and adequatelly. eLearning constitutes, therefore, a reality capable of congregating efforts towards allowing for the construction of communities involved in acquiring the necessary competences to face the challenges proposed by Globalisation. It is against this background that we present this study which aims, in its essence, at understanding the process of interaction in a collaborative distance learning environment between Classical Languages students. Being aware of the importance of promoting learning that is meaningful for the students, contents were developed representing knowledge according to Cognitive Flexibility Theory pedagogical and didactic principles. These would have to be worked on collaboratively by the participants in the study in the online discussion fórum Scaena. The work developed by the students evolved along three sessions and was integrated in the subject Educational Technology, which was part of the curriculum of the Portuguese, Latin and Greek Teacher Education Degree, 2nd Semestre, 4th year, in the academic year of 2004-05. Of a qualitative nature, the study conducted priviledged content analysis of data. For this effect the analysis software NUD*IST was used. Results revealed the occurrence of interaction patterns in all dimensions of analysis as well as the evidence of the process of flexible construction of knowledge in an oline learning platform. Finally the results obtained confirm the added value of the use of Information and Communication Technologies for Classical Studies in the educational context. Suggestions for future studies are put forward.

Techniques for batch reinforcement learning in robotics

This thesis addresses the Batch Reinforcement Learning methods in Robotics. This sub-class of Reinforcement Learning has shown promising results and has been the focus of recent research. Three contributions are proposed that aim to extend the state-of-art methods allowing for a faster and more stable learning process, such as required for learning in Robotics. The Q-learning update-rule is widely applied, since it allows to learn without the presence of a model of the environment. However, this update-rule is transition-based and does not take advantage of the underlying episodic structure of collected batch of interactions. The Q-Batch update-rule is proposed in this thesis, to process experiencies along the trajectories collected in the interaction phase. This allows a faster propagation of obtained rewards and penalties, resulting in faster and more robust learning. Non-parametric function approximations are explored, such as Gaussian Processes. This type of approximators allows to encode prior knowledge about the latent function, in the form of kernels, providing a higher level of exibility and accuracy. The application of Gaussian Processes in Batch Reinforcement Learning presented a higher performance in learning tasks than other function approximations used in the literature. Lastly, in order to extract more information from the experiences collected by the agent, model-learning techniques are incorporated to learn the system dynamics. In this way, it is possible to augment the set of collected experiences with experiences generated through planning using the learned models. Experiments were carried out mainly in simulation, with some tests carried out in a physical robotic platform. The obtained results show that the proposed approaches are able to outperform the classical Fitted Q Iteration.