基于行为学习的改进LapSVM算法及其应用

Improved Laplacian Support Vector Machine Based on Behavioral Learning and its Application

拉普拉斯支持向量机通过流形正则项能够利用未标记数据信息进行半监督学习。但其流形正则项中的数据邻接图由于没有利用数据的标记信息而不能准确表征数据流形结构,并且热核参数的经验式选择也无法保证算法的学习性能。为此,基于人类行为认知的思想构造一种新的数据邻接图:首先设计一种能够利用数据标记信息的行为相似性边权值,然后所提出的局部视角距离不仅反映邻域结构特性而且克服了热核参数选择的问题。在公共数据集上的实验验证了所提出算法的性能,最后将之应用于辐射源个体识别。

Laplacian support vector machine could utilize the unlabeled samples for semi-super- vised learning through the manifold regularization term. But the data adjacent graph in the mani- fold regularization term couldn＇ t take advantage of the label information and the empirical set- ting of heat kernel parameter would also degrade the learning performance. Inspired by human behavioral learning theory, a novel semi-supervised learning with local behavioral similarity is proposed to solve those problems. In detail, the new edge weight with label information is intro- duced and the local view distance is also applied which can not only reflect the underlying proba- bility distribution in the neighborhood but also overcome the problem of heat kernel parameter selection. Extensive experiments on toy datasets and specific emitter datasets show the validity of the new algorithm.