Filtered Reinforcement Learning
In: European Conference on Machine Learning, Jun 2004, Pisa.
Reinforcement learning (RL) algorithms attempt to assign
the credit for rewards to the actions that contributed to the reward.
Thus far, credit assignment has been done in one of two ways: uniformly,
or using a discounting model that assigns exponentially more credit to
recent actions. This paper demonstrates an alternative approach to tem-
poral credit assignment, taking advantage of exact or approximate prior
information about correct credit assignment. In nite impulse response
(IIR) lters are used to model credit assignment information. IIR lters
generalise exponentially discounting eligibility traces to arbitrary credit
assignment models. This approach can be applied to any RL algorithm
that employs an eligibility trace. The use of IIR credit assignment lters
is explored using both the GPOMDP policy-gradient algorithm and the
Sarsa( ) temporal-di erence algorithm. A drop in bias and variance of
value or gradient estimates is demonstrated, resulting in faster conver-
gence to better policies.