Further Study of the Hybrid Learning Algorithm of the TSK Fuzzy Systems
Yushun Guo · Journal of Hangzhou Dianzi University · 2010
The consequent parameters of a TSK fuzzy logic system can always be determined by the premise parameters using least squares method,and the system can thus be seen as a system which depends on the premise parameters only.From this point of view,learning of such kind of systems can be formulated as an optimization problem with respect to the premise parameters only,with the advantage of reduced dimensions and easy convergence.The gradient descendent optimizing of this problem is equivalent to the hybrid learning algorithm with one BP iteration with respect to the premise parameters and the least squares solution to the consequent parameters.The quasi-Newton optimizing of the problem leads to an efficient Newton-type hybrid learning algorithm.The conclusion drawn in this paper is also of significance for the development of more efficient learning algorithms that exploits other advanced optimizing techniques.