Gravar-mail: Reinforcement learning in dendritic structures