A Unifying View of Linear Function Approximation in Off-Policy RL Through Matrix Splitting and Preconditioning