1. Learn
    2. /
  2. Courses
    3. /
  3. Deep Reinforcement Learning in Python
Connected

Exercise

Training the double DQN

You will now modify your code for DQN to implement double DQN.

Double DQN requires only a minimal adjustment to the DQN algorithm, but goes a long way towards solving the Q-value overestimation problem and often performs better than DQN.

Instructions

100 XP
  • Calculate the next actions for the Q-target calculation using the online_network(), making sure to obtain the right action and shape.
  • Estimate the Q-values to these actions with the target_network(), again, making sure to obtain the correct values and shape.