Deep Q-learning to balance a cart pole

In the previous sections, we learned about taking an action based on q-table values. However, arriving at an optimal value is time-consuming, as the agent would have to play multiple times to arrive at the optimal q-table.

In this section, we will learn about using a neural network so that we can arrive at the optimal values faster than what we achieved when we used Q-learning.

Getting ready

For this exercise, we will register the cart-pole environment where the possible actions are to move either right or left so that we balance the pole. Additionally the cart position, cart velocity, pole angle, and pole velocity at the tip is the information we have about the states.

The rules of this...