Agent Reinforcement Learning Method Based on Credit Assignment Function
Weisheng Xu · Journal of Tongji University · 2003
The paper analyzes the principle of reinforcement learning and standard Q-Learning,and presents a new learning method based on credit assignment function called credit assignment reinforcement learning,which adjusts the coefficient of knowledge and selects proper credit assignment function to modify selection strategy of reinforcement learning.Simulation experiment shows the effectiveness of this method comparing with the stand Q-Learning method in learning rate,can improve the learning rate of standard Q-Learning and have good leaning performance.