Model-BasedMulti-AgentRLinZero-SumMarkov GameswithNear-OptimalSampleComplexity

Model-based reinforcement learning (RL), which finds an optimal policy using anempirical model, has long been recognized asone ofthe cornerstones ofRL.