الفهرس | يوجد فقط 14 صفحة متاحة للعرض العام |
المستخلص This thesis presents a proposed fuzzy neural model based on mutual subsethood measure that handles both of numeric and linguistic inputs simultaneously. Fuzzy rule- based knowledge is interpreted into network architecture. Connections in the network are represented by Bell fuzzy sets. The firing degrees of the fuzzy IF-Then rules in the proposed model are obtained based on fuzzy mutual subsethood similarity measure, which is computed neither approximately nor numerically. It is computed by an exact formula. A supervised learning procedure based on gradient descent is employed to train the network. The proposed model is considered to be a fuzzy neural model with high nonlinear capabilities. The proposed model is featured by its low cost of computations since it utilizes only one generalized analytical formula for computing firing degree of the fuzzy If-Then rule assigned to each fuzzy neuron. Conversely, other fuzzy neural models suffer of high cost of computations since they utilize several analytical formulas for different cases that results in long and tedious calculation and time consuming |