Maximum-entropy-based multiple kernel fuzzy c-means clustering algorithm

For the single kernel based clustering methods, the selection of kernel parameters largely affects the clustering results. To address this issue, a new multiple kernel fuzzy c-means clustering algorithm is proposed, in which the maximum entropy method is used to regularize the kernel weights and decide the important kernels. A new objective function is developed to simultaneously minimize the within cluster dispersion in the kernel space and maximize the kernel-weight-entropy. Thus, the optimal clustering results have been yielded and the important kernels are extracted according to the optimal assignment of kernel weights. Experiments on synthetic `nonspherical´ shaped datasets have demonstrated the efficiency and superiority of the presented algorithms.