Due to the advances in ANN (artificial neural networks), it isalready possible to use this type of intelligent system in real-timeapplications. Among the applications that would benefit from thistechnology is process control. The long time generally required fortraining ANN has been a critical problem for the utilization of thistechnology in real-time. The generalized delta rule with backward errorpropagation (backpropagation) has been the most used training algorithmfor ANN in control applications. However, despite many attempts toimprove the performance of this learning algorithm, there is still noefficient and reliable method to train a multilayer perception. The mainproblem is high nonlinearity of the error function. This paper presentsan alternative method for load information into a multilayer perception.The idea is to use successive quadratic approximations for the errorfunction, until getting into the proximity of the global minimum. Thistechnique is applied to a control system in order to obtain adaptivebehaviour
展开▼
