長遅延報酬対象向け強化学習手法 = Practical reinforcement learning for long delayed reward problems