Hands-On Intelligent Agents with OpenAI Gym

Implementing the n-step return calculation

If we pause for a moment and analyze what is happening, you might see that we are probably not making full use of the 5-step long trajectory. With access to the information from the agent's 5-step long trajectory starting from state , we only ended up learning one new piece of information, which is all about to update the actor and the critic (). We can actually make the learning process more efficient ...