集成极限学习机在手写体数字识别中的应用  

Integration Extream Leanring Machine Used in Handwritten Numeral Recognition

作　　者：刘聪[1] 张新英[1] 

机构地区：[1]中原工学院信息商务学院,河南郑州450007

出　　处：《河南科技》2016年第3期27-28,共2页Henan Science and Technology

摘　　要：本文提出了一种ELM_Adaboost算法:把ELM(Extream leanring machine)作为弱分类器,训练ELM预测样本输出,通过Adaboost算法得到多个ELM弱分类器组成的强分类器。利用MATLAB GUI编写一套手写体数字识别系统,采集大量的学习样本,提取数字样本的"十三点网格特征",使用ELM_Adaboost建立分类器。结果表明,该算法能够达到98.5%的识别率,为手写体数字识别提供指导性作用。An algorithm called ELM_Adaboost was proposed in this paper. F irstly, taking ELM （extreamleanring machine） as a weak classifier, and then training ELM algorithm and predictting the output of samples;F inally, obtaining a strong classifiers by the p lu ra lity of the ELM weak classifiers through the adaboostalgorithm. A set of handwritten numeral recognition system was composed by MATLAB GUI, and alarge number of learning samples were collected, the Thirteen point grid feature of the digital sample wasextracted,ELM-Adaboost was used to build the classifier. The results showed that the algorithm couldachieve a recognition rate of 98.5%, which provides a guiding role for handwritten numeral recognition.

关 键 词：极限学习机 集成学习 手写体数字识别 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]