智能体通过学习最优决策来解决其决策问题.激励学习方法是智能体通过与其所处的环境交互来改进它自身的行为.Markov决策过程(MDP)模型是求解激励学习问题的一般框架,瞬时差分TD(λ)是在MDP模型下与策略相关的学习值函数的一种算法.一般情况下,智能体必须记住其所有的值函数的值,当状态空间非常大时,这种记忆的量是大得惊人的.为了解决这个问题,给出了一种遗忘算法,这种算法把心理学的遗忘准则引入到了激励学习之中.利用遗忘算法,可以解决智能体在大状态空间中的激励学习问题. The agent solves its decision-making problem by learning the optimal decision-making method, which is that the agent can improve its own behavior by interacting with the environment it is in. The Markov decision-making process (MDP) model is a general framework for solving incentive learning problems, Difference TD (λ) is an algorithm of strategy-related learning value function under MDP model. In general, an agent must remember the value of all its value functions. When the state space is very large, this kind of memory In order to solve this problem, a forgetting algorithm is proposed, which introduces the forgetting criteria of psychology into the motivational learning.Using the algorithm of forgetting, we can solve the problem of the agent in the large state space Encourage learning problems.