基于分类超平面的非线性集成学习机

Nonlinearly assembling learning machine based on separating hyperplane

针对支持向量机面临的大规模数据分类问题,提出基于分类超平面的非线性集成学习机NALM。该方法借鉴管理学中协同管理的思想,将大规模数据分成规模较小的子集,然后分别在子集上运行分类超平面算法,最后将各子集上的分类结果进行非线性集成得到最终的分类结果。该方法不仅继承了分类超平面的优点,而且还将分类超平面的适用范围从小规模数据扩展到中大规模数据,从线性空间推广到Hilbert核空间。若干数据集上的实验表明:NALM能以较少的支持向量来解决大规模样本分类问题。

Inspired by collaborative management, this paper proposed nonlinearly assembling learning machine based on sepa- rating hyperplane （NALM） to solve the problems of large-scale datasets classification in support vector machine （SVM）. In NALM, the original datasets were firstly divided into several subsets. After running the separating hyperplane （SH） algorithm on each subset, the final classification results were obtained by nonlinearly assembling each result from each subset. NALM extended the usage of SH from small scale datasets to medium and large scale datasets and from linear space to Hilbert kernel soace. Experiments on several datasets verify the effectiveness of NALM.