摘要
为解决不同类型的优化问题的算法选择,将该问题视为一个分类任务,提出基于深度学习的算法选择框架,通过建立产生问题实例样本的基准问题生成器,利用卷积神经网络模型实现问题实例在人工蜂群(ABC)、复杂差分(CoDE)和协方差自适应进化策略(CMA-ES)3种算法上的自动选择,并对算法选择问题进行实验验证,结果表明基于卷积神经网络的算法选择模型预测准确率能够达到90%,能够有效解决算法选择问题.
In order to solve the algorithm selection problem for different types of optimization problems,the problem is regarded as a classification task,and an algorithm selection framework based on deep learning is proposed.The automatic selection of three algorithms,artificial bee colony(ABC),complex difference(CoDE)and covariance adaptive evolution strategy(CMA-ES),was used as an example,and the algorithm selection problem was experimentally verified.The prediction accuracy of the algorithm selection model can reach 90%,which can effectively solve the algorithm selection problem.
作者
林秀丽
李均利
田竟民
程小帆
LIN Xiuli;LI Junli;TIAN Jingmin;CHENG Xiaofan(School of Computer Science,Sichuan Normal University,Chengdu 610101,Sichuan)
出处
《四川师范大学学报(自然科学版)》
CAS
2022年第6期830-838,共9页
Journal of Sichuan Normal University(Natural Science)
基金
国家自然科学基金(62002249)。
关键词
算法选择
深度学习
分类
卷积神经网络
algorithm selection
deep learning
classification
convolutional neural network
作者简介
通信作者:李均利(1972-),男,教授,主要从事智能计算、图像处理复杂网络和目标跟踪的研究,E-mail:li.junli@vip.163.com。