Fast Affinity Propagation by Cell-based Indexing

Abstract

Affinity Propagation is one of the fundamental clustering algorithms used in various Web-based systems and applications. Although Affinity Propagation finds highly accurate clusters, it is computationally expensive to apply Affinity Propagation to a large dataset since it requires iterative computations for all possible pairs of data objects in the dataset. To address the aforementioned issue, this paper presents efficient Affinity Propagation algorithms, namely \textit{C-AP}. In order to increase the clustering speed, C-AP employs \textit{cell-based index} to reduce the number of the computed data object pairs in the clustering procedure. By using the cell-based index, C-AP efficiently detects unnecessary pairs, which do not contribute to its clustering result. For further reducing the computation time, we also present an extension of our algorithm named \textit{Parallel C-AP} that utilizes thread-parallelization techniques. As a result, C-AP and Parallel C-AP detects the same clusters as those of Affinity Propagation with much shorter computation time. Extensive evaluations demonstrate the performance superiority of our proposed algorithms over the state-of-the-art algorithms.