Implementing a DQN and analyzing its performance solving the cartpole and lunarlander problems from OpenAI Gym