#learning loop #feedback loop #markov decision process #mdp #state-action-reward