Regresar

Reinforcement learning-based tuning algorithm applied to fuzzy identification

Abstract:

In on-line applications, reinforcement learning based algorithms allow to take into account the environment information in order to propose an action policy for the overall optimization objectives. In this work, it is presented a learning algorithm based on reinforcement learning and temporal differences allowing the on-line parameters adjustment for identification tasks. As a consequence, the reinforcement signal is generically defined in order to minimize the temporal difference. © Springer-Verlag Berlin Heidelberg 2006.