1: \begin{abstract}

2:   In this paper we propose a corrected semi-proximal ADMM (alternating direction method of multipliers)

3:   for the general $p$-block $(p\!\ge 3)$ convex optimization problems with linear constraints,

4:   aiming to resolve the dilemma that almost all the existing modified versions of

5:   the directly extended ADMM, although with convergent guarantee, often perform substantially

6:   worse than the directly extended ADMM itself with no convergent guarantee.

7:   Specifically, in each iteration, we use the multi-block semi-proximal ADMM with step-size at least

8:   $1$ as the prediction step to generate a good prediction point, and then make correction

9:   as small as possible for the middle $(p\!-\!2)$ blocks of the prediction point. Among others,

10:   the step-size of the multi-block semi-proximal ADMM is adaptively determined by the infeasibility

11:   ratio made up by the current semi-proximal ADMM step for the one yielded by the last correction step.

12:   For the proposed corrected semi-proximal ADMM, we establish the global convergence results

13:   under a mild assumption, and apply it to the important class of doubly nonnegative semidefinite

14:   programming (DNN-SDP) problems with many linear equality and/or inequality constraints.

15:   Our extensive numerical tests show that the corrected semi-proximal ADMM is superior to

16:   the directly extended ADMM with step-size $\tau=1.618$ and the multi-block ADMM with

17:   Gaussian back substitution \cite{HTY12,HY13}. It requires the least number of iterations for

18:   $70\%$ test instances within the comparable computing time with that of the directly extended ADMM,

19:   and for about $40\%$ tested problems, its number of iterations is only $67\%$ that of

20:   the multi-block ADMM with Gaussian back substitution \cite{HTY12,HY13}.

21:   \end{abstract}