摘要
为克服符号回归问题经典算法具有搜索时间过长和容易陷入局部最优的缺点,提出一种基于蒙特卡洛树搜索的符号回归算法。将符号空间划分为模型空间和系数空间;在深度策略网络指导下通过蒙特卡洛树搜索实现在模型空间内寻找合适数据集特征的公式模型;在此基础上,使用粒子群算法搜索公式模型下的系数空间,得到适应度最高的公式。实验结果表明,与GP算法相比,该算法具有适应度值更低、不易陷入局部最优的特点。
To overcome the shortcomings that symbol regression algorithm shows long search time and it is easy to fall into local optimum,a symbol regression algorithm based on Monte Carlo tree search was proposed.The symbol space was divided into model space and coefficient space.Under the guidance of deep policy network,the Monte Carlo tree search was used to look for a formula model for finding suitable dataset features in the model space.On this basis,the particle swarm algorithm was used to search the coefficient space under this formula model.Experimental results show that,compared with GP algorithm,the algorithm has lower fitness value,and it is hard to fall into local optimum solutions.
作者
鲁强
张洋
LU Qiang;ZHANG Yang(Beijing Key Lab of Petroleum Data Mining,China University of Petroleum,Beijing 102249,China;College of Geophysics and Information Engineering,China University of Petroleum,Beijing 102249,China)
出处
《计算机工程与设计》
北大核心
2020年第8期2158-2164,共7页
Computer Engineering and Design
基金
国家自然科学基金项目(61402532)
中国石油大学(北京)青年基础科研基金项目(01JB0415)
国家科技重大专项基金项目(2017ZX05018-005)。
关键词
符号回归
深度策略网络
蒙特卡洛树搜索
粒子群算法
卷积神经网络
循环神经网络
symbolic regression
deep policy network
Monte Carlo tree search
particle swarm optimization
convolutional neural network
recurrent neural network