摘要
由于传统的多核聚类算法忽略了局部密度和过度限制学习最优核的表示能力,提出了一种自适应局部核的最优邻域多核聚类算法。首先通过选择不同数目的邻域构造自适应局部核,邻域之间的相似度由一个预定义的阈值下界进行度量。然后将构造的自适应局部核应用于多核聚类模型,同时放宽了最优核的刚性约束。最后在6个具基准数据集上验证了提出聚类算法的优越性和有效性。
Since the traditional multiple kernel clustering algorithm ignores the local density and over limits the representation ability of learning optimal kernel,an optimal neighborhood multiple kernel clustering algorithm based on adaptive local kernel is put forward in this paper.Firstly,an adaptive local kernel is constructed by selecting different number of neighbors,and the similarity between neighbors is measured by a predefined lower bound of threshold.Then,the adaptive local kernel is applied to the multiple kernel clustering model,and the rigid constraint of the neighborhood optimal kernel is loosened.Finally,the superiority and effectiveness of the proposed clustering algorithm are verified on six benchmark data sets.
作者
俞磊
朱铮
蒋超
肖爽
YU Lei;ZHU Zheng;JIANG Chao;XIAO Shuang(Electric Power Research Institute,State Grid Shanghai Municipal Electric Power Company,Shanghai 200051,China)
出处
《控制工程》
CSCD
北大核心
2022年第1期182-192,共11页
Control Engineering of China
基金
国网上海市电力公司科技项目(520940170023)。
关键词
多核聚类算法
局部密度
自适应
刚性约束
Multiple kernel clustering algorithm
local density
self-adaption
rigid constraint
