tfzee

It did not converge for me and it was very slow. So i did some changes it also improves performance it solves cartpole after 30 episodes Changes to hyperparameters actual...

With the current implementation the t parameter does not have an effect on the probabilities. And other implementations also use power. (for example https://github.com/werner-duvaud/muzero-general orhttps://github.com/Zeta36/muzero)

In the paper for wgan-gp you linked, they state on page 5 "**No critic batch normalization** .... In particular, we recommend layer normalization [3] as a drop-in replacement for batch...