Numerical Integration Based on Evolutionary Functional Network

In this paper, a novel evolutionary functional network (EFN) is introduced. Firstly the generalized basis function is introduced, and then genetic programming is improved by changing the objects and structure of encoding. Sequences of generalized basis functions acts as individuals, general tree structure is used to encode them. Least square method (LSM) is used to design fitness function and by a number of evolutions, the optimum approximated model is achieved. The algorithm is used to compute the numerical integrals of all kinds of functions. Finally, results of 5 experiments show that this algorithm is effectively feasible and more accurate.