Library Open Repository
Residual Reinforcement Learning using Neural Networks
Woolford, E (2005) Residual Reinforcement Learning using Neural Networks. Honours thesis, University of Tasmania. (Unpublished)
ewFINAL_with_beginning_pages.pdf | Download (605kB)
Available under University of Tasmania Standard License.
A number of reinforcement learning algorithms have been developed that are guaranteed to converge to an optimal solution for look-up tables. However, it has also been shown that these algorithms become unstable when used directly with a function approximation system. A new class of algorithms developed by Baird (1995) were created to handle the problem that direct algorithms have with function approximation systems. This thesis focused on extending Baird's work further by comparing the performance of the residual algorithm against direct application of the Temporal Difference learning algorithm. Four benchmark experiments were used to test each algorithm with various values of lambda and alpha over a period of twenty trials. Overall it was shown that the residual algorithm outperformed direct application of the TD learning algorithm on all four experiments.
|Keywords:||reinforcement learning, neural networks|
|Publisher:||Honours thesis, University of Tasmania|
|Date Deposited:||07 Jan 2006|
|Last Modified:||18 Nov 2014 03:10|
|Item Statistics:||View statistics for this item|
Repository Staff Only (login required)
|Item Control Page|