Talk:Policy gradient method

REINFORCE algorithm

i would erase the index $\pi _{\theta }$ subscript in the expectation : $$\nabla _{\theta }J(\theta )=\mathbb {E} \left[\sum _{j\in 0:T}\nabla _{\theta }\ln \pi _{\theta }(A_{j}|S_{j})\sum _{i\in j:T}(\gamma ^{i}R_{i}){\Big |}S_{0}=s_{0}\right]$$ Do you agree ? Thanks ! Hector (talk) 15:07, 4 February 2025 (UTC)