1: \begin{abstract}

2: Operator splitting methods solve composite optimization

3: problems by breaking them into smaller sub-problems that can be solved

4: sequentially or in parallel. In this paper, we propose a unified

5: framework for certifying both linear and sublinear convergence rates

6: for three-operator splitting (TOS) method under a variety of

7: assumptions about the objective function. By viewing the algorithm as

8: a dynamical system with feedback uncertainty (the oracle model), we

9: leverage robust control theory to analyze the worst-case

10: performance of the algorithm using matrix inequalities. We then show

11: how these matrix inequalities can be used to verify sublinear/linear convergence of the TOS algorithm and guide the search for

12: selecting the parameters of the algorithm (both symbolically and

13: numerically) for optimal worst-case performance. We

14: illustrate our results numerically by solving an input-constrained

15: optimal control problem.

16: 	\end{abstract}

17: