Please enable JavaScript.
Coggle requires JavaScript to display documents.
ppo - Coggle Diagram
ppo
bugs
-
done != episode start - 1, for my rollout buffer
-
collect multiple games from each level, to save level creation time
but dont do hundreds per game, since means less variety of experience, more correlated experience
-
-
-
-
-
-
-
-
-
-
-
-
graphs of reward, value, gae etc
-
-
frame stacking, so value function knows velocity
-
-
-