摘要
算法选择实际上可视为一种学习任务.鉴于此,首先分析基于元学习思想的算法选择框架;然后从数据集特征和元算法两个角度对基于元学习思想的算法选择方法进行归纳总结;最后分析基于元学习思想的算法选择存在的问题,并指出未来发展方向.
The algorithm selection problem can be considered as a learning task. Therefore, the framework of algorithm selection based on meta-learning is analyzed firstly. Then the algorithm selection based on meta-learning is classified and summarized from the viewpoint of characteristics of data set and meta-algorithm. Finally, the problems of algorithm selection based on meta-learning are analyzed, and the develop directions are proposed in the future.
出处
《控制与决策》
EI
CSCD
北大核心
2014年第6期961-968,共8页
Control and Decision
基金
国家自然科学基金项目(70791137)
关键词
算法选择
元学习
数据集特征
algorithm selection
meta-learning
characteristics of data set
