2 ms·
RL can forget, just start training it on a dataset that is different from what it was originally trained on.
by whatshisface 5y ago
RL can forget, just start training it on a dataset that is different from what it was originally trained on.
- burning_hamster 5y agoYou are right; I should have been more specific. RL does forget in the simplest sense, i.e. that certain weights in your model drift away if the data distribution is non-stationary. Humans seem to be a bit more targeted in their forgetting.