Transmission Policies Based on Learning by Reinforcement and Stochastic Geometry for Dynamic Cellular Networks