Self learning fuzzy models using stochastic approximation
Kupper · 1994
In this paper a self learning algorithm for fuzzy relational models is proposed. During the learning phase the degrees of possibility for the rules are adjusted such that the variance of the quadratic error is minimised. The algorithm employed is based on the stochastic approximation approach. Two numerical examples, using the gas furnace data from Box and Jenkins (1970), and a nonlinear discrete time system equation prove the good performance of the self learning approach.>