Sum Rate Maximization in Muti-cell Muti-user Networks: An Inverse Reinforcement Learning-Based Approach