Nondeterministic Rewards and Actions

• The functions and can be viewed as first producing a probability distribution over outcomes based on and and then drawing an outcome at random according to this distribution - nondeterministic Markov decision process
• , but is not guaranteed to converge
• decaying weighted average of the current and the revised estimate
• , where
• convergence long = 1.5 million games in Tesauro's backgammon program