Sparse Interactions in Multi-Agent Reinforcement Learning by