Applications of deep reinforcement learning to urban transit network design