论文部分内容阅读
本文研究具有人机交互能力的强化学习算法。通过人机交互给出操作者对学习结果的性能评价,智能体系统能获得当前状态与目标状态距离的度量,有效地结合操作者的先验知识和专业知识,从而使智能体在状态空间中能进行更有效的搜索,简化复杂任务的学习过程。以猜数字游戏为例,利用提出的学习框架训练智能体具有猜数字的能力。实验结果表明,结合人机交互的强化学习算法大大提高了学习效率。加快了学习过程的收敛速度。
This paper studies intensive learning algorithms with human-computer interaction ability. Through the human-computer interaction, the operator’s performance evaluation of the learning result is given. The intelligent system can obtain the measure of the distance between the current state and the target state, effectively combining the priori knowledge and professional knowledge of the operator so that the agent in the state space Can conduct more effective search, simplify the learning process of complex tasks. Take guessing digital games as an example, we use the proposed learning framework to train the agent to guess the numbers. The experimental results show that the reinforcement learning algorithm combined with human-computer interaction greatly improves the learning efficiency. Speed up the convergence of the learning process.