摘要
深度强化学习是当前机器学习领域中一个非常重要的研究方向,它通过环境对学习的反馈交互从而实现端到端的学习,能够解决一些维度高且规模大的问题。深度强化学习已经取得了多项引入注目的成果,但其缺点是鲁棒性差、局部探索能力不强、梯度易受影响。化学反应算法能够快速实现全局搜索,易于并行实现,且具有良好的鲁棒性。因此,将化学反应优化算法应用于无模型的深度强化学习方法,实践证明有一定的加速效果。
Deep reinforcement learning is a very important research direction in the current field of machine learning.It realizes end-to-end learning through the feedback interaction of the environment to learning,and can solve some high-dimensional and large-scale problems.Deep reinforcement learning has achieved many attractive results,but its disadvantages are poor robustness,weak local exploration ability,and susceptible gradients.The chemical reaction algorithm can quickly realize the global search,is easy to implement in parallel,and has good robustness.Therefore,applying the chemical reaction optimization algorithm to the model-free deep reinforcement learning method has been proved to have a certain acceleration effect.
作者
欧阳陈华
魏书堤
张汛
OUYANG Chen-hua;WEI Shu-di;ZHANG Xun(Hengyang Normal University,Hengyang 421008,China)
出处
《电脑与信息技术》
2023年第1期8-10,共3页
Computer and Information Technology
基金
湖南省教育厅科学研究项目(项目编号:20C0290)
衡阳师范学院南岳学院教学改革研究项目(项目编号:NJY201908)。