Dead-ends and Secure Exploration in Reinforcement Learning - View it on GitHub