The question is, why do you need memory for Q-learning, if you can use NA, which, to put it bluntly, already keeps previous experience in itself, to predict actions ( Q[s',a'] )?
diraria
8,329 four 14 45
alex-rudenkiy alex-rudenkiy
2,799 2 12 25
|
Source: https://ru.stackoverflow.com/questions/887263/
All Articles