论文部分内容阅读
针对人工智能领域中非独立智能生物在强化学习方面出现的MDP环境单一、学习空间狭小等问题,提出一种基于改进XCS分类器(I-XCS)的非独立智能生物强化学习机制。此学习机制在原有的XCS分类能力以及在线知识基础之上,使用梯度下降相关技术构造了一个具有高稳定性、低维度特点的逼近方法,此方法存储空间要求低,可以提升智能生物学习归纳能力。实验结果证明I-XCS分类学习算法不仅能够有效地解决MDP环境单一、学习空间狭小等问题,结论在一定程度上提高了非独立智能生物在强化学习中的分析性能。
In view of the single environment of MDP and the narrow learning space of non-independent intelligent creatures in the field of artificial intelligence, a non-independent intelligent bio-reinforcement learning mechanism based on improved XCS classifier (I-XCS) is proposed. Based on the original XCS classification ability and on-line knowledge, this learning mechanism constructs an approximation method with high stability and low dimension using the technique of gradient descent. This method has the advantages of low storage space requirement and enhanced learning ability of intelligent biology . The experimental results show that the I-XCS classification learning algorithm can not only effectively solve the single MDP environment and narrow learning space, but also improve the analytical performance of non-independent intelligent creatures in enhanced learning to a certain extent.