MARRGM: Learning Framework for Multi-Agent Reinforcement Learning via Reinforcement Recommendation and Group Modification | Synapse