|国家预印本平台
首页|集成学习选择必要性分析

集成学习选择必要性分析

Research of the Neural Network Ensemble Based on Complement

中文摘要英文摘要

提高神经网络的泛化能力是神经网络研究的热点问题之一。个体网络集成后一般可以有效地提高系统泛化能力,但并不是任意一组有差异的网络进行集成,都能使泛化能力有很大的提升。集成系统泛化能力是否能够得到提高,其关键因素在于被集成的个体网络间的互补性,而非差异性。如果仅从差异性角度进行集成,一些存在互斥性即互补性较差的网络可能导致集成后的系统泛化能力不高甚至降低。因此有选择的进行网络集成是必要的。

Improving the generalization ability of neural network is a key issue of the neural network study. In general, the ensemble of individual networks can effectively improve the generalization ability。However,there are some exceptions, not all of ensembles in a group of network with diversity can improve the generation ability. It was proved that complement is the key to the generalization ability of ensemble systems. Because not all of the individual networks can complement each other well, some network existence mutual exclusion, it is necessary to do ensemble selectively.

毕咏佳

计算技术、计算机技术

选择性集成泛化精度神经网络集成互补性

Selective ensembleGeneralization accuracyNeural Network Ensembleomplementary

毕咏佳.集成学习选择必要性分析[EB/OL].(2009-04-30)[2025-08-18].http://www.paper.edu.cn/releasepaper/content/200904-921.点此复制

评论