1: \begin{abstract}

2: The framework of Integral Quadratic Constraints of Lessard et al. (2014) 

3: reduces the computation

4: of upper bounds on the convergence 

5: rate of several optimization algorithms to 

6: semi-definite programming (SDP). Followup work by

7: Nishihara et al. (2015) applies this technique to the entire family

8: of over-relaxed 

9: Alternating Direction Method of Multipliers (ADMM). 

10: Unfortunately, they only provide an

11: explicit error bound for sufficiently large values

12: of some of the parameters of the problem, leaving the computation

13: for the general case as a numerical optimization problem. 

14: %

15: In this paper we provide an exact analytical solution to this SDP

16: and obtain a general and explicit  

17: upper bound on the convergence rate of the entire family of

18: over-relaxed ADMM. 

19: %

20: Furthermore, we demonstrate

21: that it is not possible to extract from this SDP a general bound

22: better than ours.

23: %

24: We end with a few numerical illustrations of our

25: result and a comparison between the

26: convergence rate we obtain for the ADMM with known

27: convergence rates for the Gradient Descent.

28: \end{abstract}

29: