A Data-Efficient Reinforcement Learning Method Based on Local Koopman Operators | IEEE Conference Publication | IEEE Xplore