Simulation and Research for Forward Neural Networks Modeling

This paper presents analysis and experiments for Forward Neural Networks (FNN) modeling based on software simulation. First, we illustrate this modeling method in detail. Then the algorithm flow charts of four learning algorithms, including Back Propagation (BP), Modified BP (MBP), Davidon Least Square based algorithm and Damping Least Square based algorithm are given. In addition, a simulation software for training and testing FNN is developed using these algorithms. In the last, we compare and analyze the performance of these four algorithms based on simulation experimental results.