Model-Based ReinforcementLearningviaImagination withDerivedMemory