An Implementation of First and Second Order Neural Network Classification on Potential Drug Addict Repetition

Abstract

Back propagation (BP) neural network is known for its popularity and its capability in prediction and classification. BP used gradient descent (GD) method as one of the most widely used error minimization methods used to train back propagation (BP) networks. Besides its popularity BP still faces some limitation such as very slow in learning as well as easily get stuck at local minima. Many techniques have been introduced to improve BP performance. This research implements second order method together with gradient descent in order to improve its performance. The efficiency of both methods are verified and compared by means of simulations on classifying drug addict repetition. The results show that the second order methods are more reliable and significantly improves the learning performance of BP.