Computationally Efficient Reinforcement Learning: Targeted Exploration leveraging Simple Rules