Designing a neural network by a genetic algorithm with partial fitness

This paper presents a method of using the genetic algorithm (GA) with partial fitness (PF) to design a neural network for coin recognition. The method divides a chromosome in the GA into several parts, the PFs of which are evaluated for GA operations. Each part independently performs selection and crossover operations in the GA. Such a technique improves performance in learning of the GA. This paper applies the method to a rotated coin recognition problem to examine its effectiveness. The coin recognition system described consists of a preprocessor with Fourier transform and a multilayered network. The method is utilized to reduce the number of input signals, Fourier spectra, of the multilayered network. It is shown that the method is better than the conventional GA on convergence in learning and makes a smaller size network.