Hello,
when using the Monte Carlo update in 'Reinforce with Baseline' in order to update the V values, do we consider only the discounted reward obtained in the episode or do we consider the average of the obtained rewards over all the previous episodes?
Thanks in advance!
For the V-value you should always use a good estimation. Whether you do this in an online update (over many episodes) or in a batch update (over many episodes) does not matter. But yes, many episodes)
Follow the pulses of EPFL on social networks
© 2023 EPFL, all rights reserved