1: \begin{abstract}

2:   We show that the Bellman operator underlying the options framework leads to

3:   a matrix splitting, an approach traditionally used to speed up convergence of

4:   iterative solvers for large linear systems of equations. Based on standard

5:   comparison theorems for matrix splittings, we then show how the asymptotic rate of

6:   convergence varies as a function of the inherent timescales of the options.

7:   This new perspective highlights a trade-off between asymptotic performance

8:   and the cost of computation associated with building a good set of options.

9: \end{abstract}