Training Fair Deep Neural Networks by Balancing Influence