Simplifying Deep Temporal Difference Learning

A modern implementation of Deep Q-Network without target networks and replay buffers.