Based on the methods of fuzzy central clustering algorithms from unsupervised online recursive and offline learning methods, the limitation of initial sensitivity of clustering and learning of an objective function in constrained nonlinear optimum programming are analyzed. A modified offline learning approach is presented. The advantages and disadvantages of three kinds of fuzzy central clustering algorithms are compared by way of simulation. It shows that an approach proposed here not only decreases initial sensitivity of clustering but also accelerates termination learning of an objective function.
展开▼