| |
Abstract:
We present a method for automatically constructing
macro-actions from primitive actions in reinforcement learning. The
agent constructs the macros totally from scratch during the
learning process. The overall idea is to reinforce the tendency to
perform action B after action A if such a pattern of actions has
been rewarded. We test the method on a bicycle task, the Car On The
Hill Task, the Race-Track Task and some grid-world tasks. For the
bicycle task and the Race-Track the use of macro actions
approximately halves the learning time, while for the grid-world
tasks the learning time is drastically reduced by orders of
magnitude.
|
