37 Comparing Weight Generation Methods for Neural Networks Applied to the Road Pixel Identification Problem Available to Purchase

Certain types of problems are extremely difficult to solve through the use of standard computational algorithms and require the employment of alternative methods such as artificial intelligence techniques, which aim to generate a close approximate solution rather than a definite best solution. This paper concentrates on the application of these techniques to the road pixel identification problem using spectral band analysis of satellite data. In particular, a standard back-propagation neural network and a neural network using weights evolved from a genetic algorithm are used to attempt to solve the road pixel identification problem. This paper also compares the performance of the two previously mentioned techniques. This paper attempts to determine which of the two techniques is more suitable for learning the weights of the feed forward neural network as it applies to the road pixel identification problem.