I have a general question about the updating of the network/model in the PPO algorithm.

If I understand it correctly, there are multiple iterations of weight updates done on the model with data that is created from the environment (with the model before the update). Now, I think that the updates of the model weights are not correct anymore after the first iteration/optimization step, because the model weights changed and therefore the training data is outdated (since the model would now give different actions in the environment and therefore different rewards).

Basically, in the pseudo-code of the algorithm, I don't understand the line "Optimize surrogate L ... with K epochs...". If the update is done for multiple epochs, the data that is learned is outdated already after the first iteration of optimization, since the model's weights changed. In other algorithms, like A2C, there is only one optimization step done, instead of $K$ epochs.

enter image description here

Is this some form of approximation or augmentation on the data by using the data that was created by an older model for multiple iterations or am I missing something here? If yes, where was this idea first introduced or better described? And where is an (empirical) proof that this still leads to a correct weight updating?


1 Answer 1


Seems this question was asked and answered in openai/baselines github issue. The issue has been closed for a while.

Below is an answer provided by @matthiasplappert which has the most "thumbs up":

To clarify: PPO is an on-policy algorithm so you are correct that going over the same data multiple times is technically incorrect.

However, we found that PPO is actually quite okay with doing this and we still get stable convergence. This is likely due to the proximal trust region constrained that we enforce, which means that the policy cannot change that much anyway when going over the current set of transitions multiple times, making it still approximately on-policy. You can of course get rid of this but then you'll need more samples.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .