This paper analyses a simple epsilon-greedy exploration approach to train models with Deep Q-Learning algorithm to involve randomness that helps prevail the agent over conforming to a single solution. This allows the agent ...