Closed leesharkey closed 2 years ago
It should suffice to have several samples that start in the same place in a level but use different actions. E.g. a 3x3 grid where the agent takes a constant action (e.g. grid position 3,3 would have constant downright actions)
Done in f74d91a320c5a1527a7d8a92c9747e364d0229dc
Then implement the experiments in code