[Question] How was the target entropy in the discrete SAC chosen?

p-christ / Deep-Reinforcement-Learning-Algorithms-with-PyTorch

PyTorch implementations of deep reinforcement learning algorithms and environments

MIT License

5.59k stars 1.19k forks source link

[Question] How was the target entropy in the discrete SAC chosen? #88

Open aivarsoo opened 8 months ago

aivarsoo commented 8 months ago

Hello! I have a question on the discrete SAC design.

What was the reasoning for choosing the target entropy in the discrete SAC? If I understand correctly the target entropy represents the ideal entropy of the optimal policy. If so why it is -0.98 * log( 1 / |A|)?