摘要
首先给出了属性分割的有效性测试算法,它能检测基于当前分割的交互学习算法是否优越于传统的不划分属性的单一数据集的学习算法;进一步,给出了渐进式自动属性集合的划分算法。基于大规模卫星环境监测数据的实验表明,上述两种算法能对有效地分割数据集合,使学习能力得到明显提高。
A feature split algorithm is presented which can test whether this algorithm will outperform the traditional algorithms without using a feature split with a just single view. A progressive approach is given for detecting a good feature split for co-training. Experimental results indicate that the progressive method always outperforms the traditional methods in supplying the feature division.
出处
《武汉理工大学学报（信息与管理工程版）》
CAS
2005年第2期187-191,共5页
Journal of Wuhan University of Technology：Information & Management Engineering
