Parallel Implementation of Membrane Computing-Inspired Clustering Algorithm on Graphics Processing Unit

Membrane computing is a class of distributed parallel computing models, and it has been considered to deal with data clustering recently. The membrane computing-inspired clustering algorithms are called membrane clustering algorithms that use a variety of membrane systems as their key components. However, the membrane clustering algorithms were only realized in a serial algorithm form because of serial architecture of current computer. Therefore, the membrane clustering algorithms were not able to exhibit the parallel computing characteristic of membrane systems. This paper focuses on parallel implementation of membrane clustering algorithms and proposes a GPU (Graphics Processing Unit)-based parallel computing framework and parallel version of a membrane clustering algorithm. In the parallel implementation, the blocks are used to represent the cells, while threads are considered to realize the evolution-communication mechanism of objects. The comparison results on several artificial and real-life data sets demonstrate that the proposed parallel version not only ensures the clustering quality of the membrane clustering algorithm but also evidently reduce its computing time.