The design of self-organizing evolved polynomial neural networks based on learnable evolution model 3

الملخص EN

Nowadays, the development of advanced techniques of system modeling has received much attention.

Polynomial Neural Network (PNN) is a GMDH-type algorithm (Group Method of Data Handling), which is one of the useful methods for modeling nonlinear systems but PNN performance depends strongly on the number of input variables and the order of polynomial which are determined by trial and error.

In this paper, we discuss a new design methodology for polynomial neural networks PNN in the framework of Learnable Evolution Model (LEM3).

LEM3 is a new approach to evolutionary computation, which employs machine learning to guide evolutionary processes.

LEM3 is obtained better performance in shorter time in comparing with other well-known methods.

Also, LEM3 appears to be particularly suitable for solving complex optimization problems in which the fitness evaluation function is time consuming.

In this paper, we use LEM3 to search between all possible values for the number of input variables and the order of polynomial.

Evolved PNN performance is obtained by two nonlinear systems.

The experimental part of the study involves two representative time series such as Box-Jenkins gas furnace process and the Dow Jones stock index.