Q: What are the limitations of Q-learning, and how does DQN improve it?