Adaptive Affinity Propagation Clustering
自适应仿射传播聚类

Affinity propagation (AP) clustering has two limitations: it is hard to know what value of parameter ``preference' can yield optimal clustering solutions, and oscillations cannot be eliminated automatically if occur. This paper proposes an adaptive AP method to overcome these limitations, including adaptive scanning of preferences to search space for finding the optimal clustering solution, adaptive adjustment of damping factors to eliminate oscillations, and adaptive escaping from oscillations when the damping-factor adjustment technique fails. In comparison to AP, the adaptive AP has better performance on automatic oscillation elimination and finding of an optimal clustering solution. Experimental results on simulated and real data sets show that the adaptive AP is effective and its quality of clustering results is better than or equal to that of AP.