A Framework for Aggregation of Multiple Reinforcement Learning Algorithms