What is the best way of using the LSTM layer in PPO architecture? Should I use them in the first layer of both actor and critic, or use them just before the final layer of these networks? Should I feed the architecture with a stack of states (the state stacked with the k previous states)?
Not sure if you're using TensorFlow-Agents for this, but if you are, there's some nice functionality for this built out using the
ValueDistributionRnnNetwork that I've found to be very helpful. The LSTM architecture is used in both the actor and critic networks. See this page for more information - hope this helps!