====== Action Space ======

===== Large Action Space =====

  * [[https://arxiv.org/pdf/2001.08116.pdf|Q-Learning in Enourmous Action Spaces via Amortized Approximate Maximization, 2020-01]]
    * DeepMind, Volodymyr Mnih

  * action space restriction
    * masked softmax
      * https://torchcraft.github.io/TorchCraftAI/docs/bptut-rl.html
      * https://gist.github.com/kaniblu/94f3ede72d1651b087a561cf80b306ca
      * https://discuss.pytorch.org/t/apply-mask-softmax/14212/12