Content area
Full Text
Keywords Mathematical modelling, Programming and algorithm theory, Neural nets
Abstract In this research, neural network (NA and genetic algorithm (GA) are used together to design optimal NN structure. The proposed approach combines the characteristics of GA and NN to reduce the computational complexity of artificial intelligence applications in design and manufacturing. Genetic input selection approach is introduced to obtain optimal NN topology. Experimental results are given to evaluate the performance of the proposed system.
1. Introduction
Neural network (NN), also called artificial neural system, is an information processing technique which is developed to simulate the functions of a human brain. Although NN is an effective algorithm to solve engineering problems, there are only few approaches to design the network and most of them rely on iterative procedures. The design of network architecture mainly consists of the selection of the training algorithm, the determination of network layers and the number of neurons in each layer. Despite its rise to use in a great diversity of science and engineering problems, there is still no general procedure to design NN architecture. Therefore, this often causes over design or inefficient network structures especially in the case of complex problems.
In order to make NN-based applications more efficient, there is a need to improve the convergence speed and reduce the computational complexity of NN. The computational complexity and convergence speed of network are generally affected by the number of neurons in each layer since it acts poorly as model become larger and more complex. For example, it usually takes a long time to train because of its slow convergence of often used back propagation training algorithm.
Genetic algorithm (GA) is an evolutionary search algorithm based on the mechanics of natural selection and natural genetics and has capability to search large number of combinations, as there may be interdependencies and redundancies between variables (Goldberg, 1989). GA is generally used as an optimisation technique. Therefore, in this research, the advantages of NN and GA are used together in a hybrid approach to design optimal NN structure. The objective is to reduce the computational complexity and the time required to design the NN. Illustrative computer aided design and manufacturing examples are given to evaluate the results of the present approach.
2....