Which of the following is a requirement on the behavior policy b for using off-policy Monte Carlo policy evaluation? This is called the assumption of coverage.

Question

Which of the following is a requirement on the behavior policy b for using off-policy Monte Carlo policy evaluation?   This is called the assumption of coverage.

Accepted Answer

For each state s and action a, if π(a∣s)>0 then b(a∣s)>0

Answer

For each state s and action a, if b(a∣s)>0 then π(a∣s)>0

Answer

All actions have non-zero probabilities under π

Crowdly

Which of the following is a requirement on the behavior policy b for using off-p...

Хочете миттєвий доступ до всіх перевірених відповідей на elearning.aua.am?