基于相融性度量的光谱分类方法

海量天体光谱的自动分类以及从海量天体光谱中发现新类型天体或新的天文规律（知识发现）已经受到天文工作者的普遍关注。在相关文献中这两方面的研究工作都是分别进行的。文章首先提出了一种相融性度量的概念，该度量能够刻画一个样本与训练样本集融合为一体的程度。然后，在此基础上给出了一种基于相融性度量的k-近邻分类方法。该方法不仅能够实现较准确的分类，而且还具有相当好的知识发现能力。通过对活动星系与活动星系核实验表明，该方法无论对分类还是对知识发现都是非常有效的。

A Novel Spectral Classifier Based on Coherence Measure

Classification and discovery of new types of celestial bodies from voluminous celestial spectra are two important issues in(astronomy),and these two issues are treated separately in the literature to our knowledge.In the present paper,a novel coherence measure is introduced which can effectively measure the coherence of a new spectrum of unknown type with the training samples located within its neighbourhood,then a novel classifier is designed based on this coherence measure.The proposed classifier is capable of carrying out spectral classification and knowledge discovery simultaneously.In particular,it can effectively deal with the situation where different types of training spectra exist within the neighbourhood of a new spectrum,and the traditional k-nearest neighbour method usually fails to reach a correct classification.The satisfactory performance for classification and knowledge discovery has been obtained by the proposed novel classifier over active galactic nucleus(AGNs) and active galaxies(AGs) data.