Transfer Learning via Multiple Inter-Task Mappings

A. Fachantidis, I. Partalas, M. Taylor, I. Vlahavas, “Transfer Learning via Multiple Inter-Task Mappings”, Recent Advances in Reinforcement Learning, Springer Berlin Heidelberg, pp. 225-236, 2012.

Recent Advances in Reinforcement Learning, Springer Berlin Heidelberg, pp. 225-236, 2012.

In this paper we investigate using multiple mappings for transfer learning in reinforcement learning tasks. We propose two dif- ferent transfer learning algorithms that are able to manipulate multiple inter-task mappings for both model-learning and model-free reinforce- ment learning algorithms. Both algorithms incorporate mechanisms to select the appropriate mappings, helping to avoid the phenomenon of negative transfer. The proposed algorithms are evaluated in the Moun- tain Car and Keepaway domains. Experimental results show that the use of multiple inter-task mappings can signi?cantly boost the performance of transfer learning methodologies, relative to using a single mapping or learning without transfer.