Please enable JavaScript.

Coggle requires JavaScript to display documents.

Intrinsic Motivation (evaluation (montezuma, pitfall, coinRun-Platform…

- - - - feed in phi(s(t-1)) and a(t-1)
      - use recurrent network
    - - done by Brunner et al
    - - forward model outputs additional certainty value,trained to be 1, high certainty increases error/learning rate
      - can be used as intrinsic reward itself