A FAST-Based Q-Learning Algorithm
Kao‐Shing Hwang, Yuan-Pao Hsu, Hsin-Yi Lin · BiblioBoard Library Catalog (Open Research Library) · 2009
This article proposed a reinforcement learning architecture that combines ARM, a FAST? based algorithm, and Q-learning algorithm. The ARM, at the front end, is featured with multi-neuron triggering and dynamically adjusting its sensitive region as well, providing the Q-learning, at the back end, with more suitable clustered input states. In such a way, the Q-learning learns quick and stable. There are future work can be tried to enhance the Q-learning for constructing more efficient reinforcement learning architecture, such as replaces the Q-learning in our architecture with the Q()-learning [11] or SARSA [12].