Reinforcement Learning Control using Self-Organizing Map and Multi-layer FEED-Forward Neural Network