search

mobeets / sarsa-rnn

0 stars 0 forks source link

neurogym: add correctness wrapper #11

Open mobeets opened 2 years ago

mobeets commented 2 years ago

I want the observations to have whether a trial was correct or not, not just the reward amount

mobeets commented 2 years ago

Like #12, this is also trickier than it sounds. Is there a separate observation for correct and incorrect, or is it a single dimension with a signed value? Plus, needing the task to define a correct vs. incorrect action.