1: \begin{abstract}

2: In this paper, we propose an {\em inexact} multi-block ADMM-type first-order

3: method  for solving a class of high-dimensional convex composite conic optimization problems to moderate accuracy. The design of this method combines an inexact 2-block majorized semi-proximal ADMM and the recent advances in the  {inexact} symmetric Gauss-Seidel (sGS) technique

4:  for solving a multi-block convex composite quadratic programming whose objective

5: contains a nonsmooth term involving only the first

6: block-variable.

7: One distinctive feature of our proposed method (the sGS-imsPADMM) is that it only needs one cycle of an inexact sGS method, instead of an unknown number of

8: cycles,

9: to solve each of the subproblems involved.

10: With some simple and implementable error tolerance criteria, the cost for solving the subproblems can be

11:  greatly  reduced,

12: and many steps in the forward sweep of each sGS cycle can often be

13: skipped,

14: which further contributes to the efficiency of the proposed method.

15: Global convergence as well as the iteration complexity in {the non-ergodic sense is}

16: established.

17: Preliminary numerical experiments on some high-dimensional linear and convex quadratic SDP problems with a large number of

18:  linear equality and inequality constraints

19: are also provided.  The results show that for the vast majority of the tested problems,  the sGS-imsPADMM is 2 to 3 times faster than the directly extended multi-block ADMM with  the aggressive step-length of  $1.618$, which

20: is currently the benchmark

21: among first-order methods for solving multi-block linear and quadratic SDP problems though its convergence is not guaranteed.

22: \end{abstract}

23: