Agglomerative clustering via maximum incremental path integral
作者：
Highlights：
• A novel graph-structural agglomerative clustering algorithm.
• A structural descriptor, Incremental Path Integral, to define affinity of clusters.
• Treat a cluster as a dynamic system and its samples as states.
• Path Integral measures the stability of a dynamic system.
• Considerably outperform the state-of-the-art clustering algorithms.
摘要
Highlights•A novel graph-structural agglomerative clustering algorithm.•A structural descriptor, Incremental Path Integral, to define affinity of clusters.•Treat a cluster as a dynamic system and its samples as states.•Path Integral measures the stability of a dynamic system.•Considerably outperform the state-of-the-art clustering algorithms.
论文关键词：Agglomerative clustering,Path integral,Graph algorithms,Random walk
论文评审过程：Received 18 July 2012, Revised 3 April 2013, Accepted 10 April 2013, Available online 24 April 2013.
论文官网地址：https://doi.org/10.1016/j.patcog.2013.04.013
