Learning agents for federated collaborative virtual workspace
LE3 .A278 2006
2006
Shakshuki, Elhadi
Acadia University
Master of Science
Masters
Computer Science
Collaborative Virtual Workspace (CVW) is an environment for collaboration and knowledge management for temporally and geographically dispersed work teams. Federated Collaborative Virtual Workspace (FCVW) is an extension of CVW that supports multiple servers to connect and collaborate with one another. Today's Collaborative Virtual Environments (CVEs) lacks software agents that can interact with users to learn their preferences so as to facilitate the user toward preferential solutions. Recently, three types of software agents have been created in FCVW including observer agents, garbage collector agents and information retrieval agents. Although these agents are capable of performing several important tasks for the user, they lack learning capabilities. An agent is developed with the ability to monitor and learn user preferences to predict future actions of the user. This agent is equipped with two learning techniques, namely; Genetic Algorithm (GA) and Reinforcement Learning Algorithm (RA). The agent is a Learning Agent and has the capability to switch automatically between the two learning techniques. The main objective of the learning agent is to monitor the file selection of the user in FCVW rooms and learn his or her preferences to predict the files of his or her interest. The learning agent is implemented and demonstrated in FCVW. To evaluate its capabilities, a performance measure is performed on the accuracy of predictions of files, training time, processing time and memory utilization.
The author retains copyright in this thesis. Any substantial copying or any other actions that exceed fair dealing or other exceptions in the Copyright Act require the permission of the author.
https://scholar.acadiau.ca/islandora/object/theses:2721