Please use this identifier to cite or link to this item:
Title: Multi-task learning and transfer: The effect of algorithm representation
Authors: Lane, PCR
Gobet, F
Keywords: Learning;Model;Theory development;Learning algorithm;Concept formation;CHREST;Connectionism;Mathematical model
Issue Date: 2005
Publisher: Proceedings of the ICML-2005 Workshop on Meta-learning
Citation: Proceedings of the ICML-2005 Workshop on Meta-learning.
Abstract: Exploring multiple classes of learning algorithms for those algorithms which perform best in multiple tasks is a complex problem of multiple-criteria optimisation. We use a genetic algorithm to locate sets of models which are not outperformed on all of the tasks. The genetic algorithm develops a population of multiple types of learning algorithms, with competition between individuals of different types. We find that inherent differences in the convergence time and performance levels of the different algorithms leads to misleading population effects. We explore the role that the algorithm representation and initial population has on task performance. Our findings suggest that separating the representation of different algorithms is beneficial in enhancing performance. Also, initial seeding is required to avoid premature convergence to non-optimal classes of algorithms.
Appears in Collections:Psychology
Dept of Life Sciences Research Papers

Files in This Item:
File Description SizeFormat 
lane-icml05.pdf133.69 kBAdobe PDFView/Open

Items in BURA are protected by copyright, with all rights reserved, unless otherwise indicated.