选择性集成学习算法综述

摘要

In recent years, ensemble learning has received lots of attention in machine learning due to its potential to significantly improve the generalization capability of a learning system. With increasing number of ensemble members, however, the prediction speed of an ensemble machine decreases significantly and its storage need increases quickly. The aim of selective ensemble learning is to further improve the prediction accuracy of an ensemble machine, to enhance its prediction speed as well as to decrease its storage need. This paper presents a detailed review of the current selective ensemble learning algorithms and categorizes them into different classes according to their utilized selection strategy. Meanwhile, the main characteristics of each representative algorithm are studied. Finally, the future research directions of selective ensemble learning are discussed.%集成学习因其能显著提高一个学习系统的泛化能力而得到了机器学习界的广泛关注,但随着基学习机数目的增多,集成学习机的预测速度明显下降,其所需的存储空间也迅速增加.选择性集成学习的主要目的是进一步改善集成学习机的预测效果,提高集成学习机的预测速度,并降低其存储需求.该文对现有的选择性集成学习算法进行了详细综述,按照算法采用的选择策略对其进行了分类,并分析了各种算法的主要特点,最后对选择性集成学习在将来的可能研究方向进行了探讨.