Integration of Efficient Deep Q-Network Techniques Into QT-Opt Reinforcement Learning Structure