Minimum Enclosing Ball-based Learner Independent Knowledge Transfer for Correlated Multi-task Learning
Institution: | AUT University |
---|---|
Department: | |
Year: | 2011 |
Keywords: | Multi-task Learning; Knowledge Transfer; Correlated multi-task learning; Minimum Enclosing Ball; Machine Learning; Knowledge Sharing; Learner Independence |
Record ID: | 1303290 |
Full text PDF: | http://hdl.handle.net/10292/1120 |
Multi-Task Learning (MTL), as opposed to Single Task Learning (STL), has become a hot topic in machine learning research. For many real world problems in application areas such as medical decision making, pattern recognition, and finance forecasting – MTL has shown significant advantage to STL because of its ability to facilitate knowledge sharing between tasks. This thesis presents our recent studies on Knowledge Transfer (KT) – the process of transferring knowledge from one task to another, which is at the core of MTL. The novelly proposed KT algorithm for correlation multi-task machine learning adapts learner independence into MTL, thus empowering any ordinary classifier for MTL.