METHOD FOR TRANSFORMING A RADIALLY BASED NEURAL NETWORK INTO A FUZZY INFERENCE SYSTEM

Abstract

METHOD FOR TRANSFORMING A RADIAL BASED NEURAL NETWORK INTO A FUZZY INFERENCE SYSTEM The present invention relates to a method for transforming a radial basis neural network (RBFN) into a fuzzy inference system. This method comprises a step of dividing the network into distinct RBFN subnetworks and then separating (120) each of these subnetworks. A layer of inference nodes is then inserted (130) at the output of the subnetworks, each inference node applying an inference rule, and then the subnetworks with this inference layer are subjected to a training phase (140) on a training dataset. The activation functions of the neurons in the intermediate layer are then degraded (150), those relating to neurons connected to the same node of the input layer being degraded into membership functions representing a partitioning of the input variable space.A fuzzy inference system functionally equivalent to RBFN is finally generated (160) from the membership functions and inference rules previously adjointed. Figure for the abstract: Figure 1.

Claims

8. Method according to claim 6 or 7, characterized in that each inference rule is expressed in the form of a piecewise multivariate polynomial whose variables are input variables of the RBFN network, said method comprising a calculation step (370) of the coefficients of the multivariate polynomial from the centers of the radial activation functions of the RBFN network.

9. Method according to claim 8, characterized in that it comprises a memory storage (380) of the coefficients of the piecewise multivariate polynomials expressing the inference rules of the fuzzy inference system as well as the centers of the radial activation functions of the RBFN network.

10. Embedded system comprising sensors and implementing a fuzzy inference system, characterized in that the fuzzy inference system is generated by a method according to one of the preceding claims, the input variables being physical measurements provided by the sensors of the embedded system and the output variables being control parameters of said system. S-shaped division of distinct RBFN subnetworks 110 f Separation of each of the RBFN subnetworks 120 / f Insertion of an inference node layer 130 l Learning of the subnetworks equipped with inference layers 140 f Degradation of the activation functions of neurons 150 f Obtaining a functionally equivalent FIS to the initial RBFN 160 220 230

Images