Efficient memory models using memory monoids can improve sample efficiency and simplify implementation in recurrent reinforcement learning.
