Non-independent Intelligent Creatures Reinforcement Learning Mechanism Research Based on I-XCS
PengJian Xi, Jianxiong Tan · Advances in intelligent systems research/Advances in Intelligent Systems Research · 2015
In order to solve many problems of reinforcement learning of Non-independent intelligent creatures in artificial intelligence, such as the single MDP environment and narrow learning space.This paper designed an Non-independent intelligent creatures reinforcement learning mechanism based on the Improved XCS classifier.This learning mechanism based on the original XCS classification capabilities and online knowledge, it constructs a high-stability, low-dimensional approximation method by using the gradient descent related technologies.This method has low-storage ability and enhances the inductive learning ability of intelligent creatures.Simulation experiment results show that the I-XCS classification learning algorithm not only can efficiently solve MDP environment issues such as single, narrow space, but also to a certain extent improved the analysis of non-independent intelligent creatures in reinforcement learning performance.