Skip to content
ml-recall

#surrogate-objective

1 question from Policy Gradient & Actor-Critic

Timed test

1 question · 1 min · score and explanations at the end

Take a test
Q 1 / 1Score 0 / 0

Question 1: True or false

mediumAI-checkedPolicy Gradient & Actor-Critic

In a vanilla policy-gradient implementation the "loss" is -(logp * weights).mean() over a batch of data from the current policy. If this loss keeps decreasing during training, the policy's expected return must be improving.

Choose one option