04ReinforcementLearning3.4, Log-likelihood trick: from batch to online

views comments

When going from batch to online, care has to be taken to arrive at the statistical weight of online sampling. The log-likelhood trick solves this issue.