[2411.07591] Overcoming the Curse of Dimensionality in Reinforcement Learning Through Approximate Factorization