人脑分类机理的构造性学习方法

Constructive Learning Method Based on Classification Mechanism of Human Brain

构造性学习(CML)算法训练分类器对有些样本会有"拒认状态",构造性学习算法中对这一状况的处理使用就近原则,然而,这种方法无法体现数据之间的联系.为了能更好地体现数据间的联系,提出了人脑分类机理的构造性学习方法(HB-CML).在测试阶段,把测试样本、训练样本都考虑进来,利用人脑对数据的自动分类机理,对"拒认状态"样本进行分类标记.同时,选取UCI数据集进行实验.结果表明:与CML算法相比,该方法的分类更为有效.

＂ Refusing to be classified＂ test examples will be produced using Constructive Machine Learning （CML） algorithm and the examples will be labeled according to principle of proximity, however the connections between data is ignored. So a constructive learning method based on human brain algorithm （ HB - CML） is designed to reflect the connections between labeled and unlabeled samples. During the testing phase, the＂ refusing to be classified＂ test examples are labeled by automatic data classification mechanism of human brain using labeled and unlabeled samples. At the same time, experiment is conducted on UCI data set and results show that the algorithm is more effective than the CML algorithm.