A Java Framework for Reinforcement Learning Experiments

An interface between Piqle and RL_glue2 is directly available from the Sourceforge download page.

To use it inside the rl2008 competition software :

At this time, only discrete actions can be handled.

Three example agents are provided :

The learner seems to be very sensitive to initial parameters, so please if you find interesting parameter settings, let us know about them !

For further information about learners and their parameters, please refer to the documentation, or ask me directly : Francesco.De-Comite at