In PPO (or TD3), how can you both determine the minimum and maximum action the agent can take (for example, between 0 and 1) and also make sure that all the actions sum up to 1?

In Python, I can use space.box to restrict the action space and I can use the softmax to make sure the actions sum to 1. But they do not work at the same time, and so have to be done successively, which means one of the two restrictions will be broken.



You must log in to answer this question.

Browse other questions tagged .