Long Ji Lin Self-Improving Reactive Agents Based On Reinforcement Learning, Planning and Teaching. 293-321 1992 8 Machine Learning db/journals/ml/ml8.html#Lin92