The purpose of cluster analysis is to classify elements according to their similarity. Its applications range from astronomy to bioinformatics and pattern recognition. Our method is based on the idea that the data with similar distribution form a hyper-ball and the adjacent hyper-balls form a cluster. Based on the cognitive law of "large scale first", this method can identify clusters without considering shape in a simple and non-parametric way. Experimental results on several datasets demonstrate the effectiveness of the algorithm.
翻译:集束分析的目的是根据相似性对各元素进行分类,其应用范围从天文学到生物信息学和模式识别。我们的方法依据的理念是,分布相似的数据形成超球,相邻的超球形成一个集群。根据“大规模第一”的认知法,这种方法可以在不考虑简单和非参数形状的情况下识别各组。若干数据集的实验结果显示了算法的有效性。