Membrane Clustering Algorithms And Their Application In Image Segmentation

Clustering is an unsupervised learning algorithm in machine learning.It has been successfully applied in many fields,such as data mining,machine learning,image processing and biology.Clustering describes the process of grouping data into classes or clusters such that the data in each cluster share a high degree of similarity while being very dissimilar to data from other clusters.However,the traditional clustering algorithms have the following disadvantages:(1)it is sensitive to initial center,(2)it is easy to fall into local optimization,and(3)it can not well deal with the data sets with non-spherical cluster boundaries.Membrane computing is a kind of distributed parallel computing models which are inspired by the mechanism of biological cells.In order to overcome the shortcomings of the clustering algorithms,this paper introduces membrane computing into the clustering algorithms,and develops several new membrane clustering algorithms,and then discusses their applications.The innovative achievements of this paper can be summarized as follows:(1)A kernel-based membrane clustering algorithm is proposed.This algorithm maps data to high dimensional space by nonlinear mapping,and the membrane clustering algorithm is used for clustering.For this reason,a tissue P system is designed to find the optimal cluster centers.Because of the introduction of the kernel method,the algorithm can deal with the nonspherical boundary data.(2)A kernel-based fuzzy membrane algorithm with spatial constraints is proposed for image segmentation.For image pixel clustering,image data is mapped to a high-dimensional feature space,and then a tissue P system is designed for fuzzy clustering to find the optimal cluster centers,and the objective function containing spatial constraints is considered.The experimental results show that the proposed algorithm is feasible and close to the results of artificial segmentation.(3)A variant of numerical P systems is proposed,called stochastic numerical P systems.Compared with usual numerical P system,the stochastic numerical P systems introduces a stochastic production function and repartitional protocol as an evolutionary program.Experimental results on benchmark datasets demonstrate its effectiveness in dealing with clustering problems.