A performances study of enhanced bp algorithms on aircraft image classification

BP is by far the most widely used algorithm to train MLPs for pattern recognition and other similar tasks. However it is stigmatized with the problems of low convergence, instability and overfitting. In addition, the optimal values of the learning rate, momentum, the number of hidden layers and its dimension are obtained through trial and error method. In this work, we evaluate eleven (11) enhanced BP algorithms in classifying aircraft images. The image is represented using a set of Zernike Moment Invariants.