Beta. Content is under active construction and has not been peer-reviewed. Report errors on GitHub.Disclaimer

Policy Gradient Theorem

3 questionsDifficulty 6-7View topic

Intermediate

0 / 3

1 intermediate2 advancedAdapts to your performance

1 / 3

intermediate (6/10)conceptual

In the REINFORCE algorithm, the policy gradient is

\nabla_{θ} J = E [\sum_{t} \nabla_{θ} lo g π_{θ} (a_{t} ∣ s_{t}) G_{t}]

. Why is subtracting a baseline

b (s_{t})

from

G_{t}

useful even though it does not change the expected gradient?