1: \begin{abstract}

2: We consider a primal-dual algorithm for minimizing $f(\vx)+h(\vA\vx)$ with differentiable $f$. 

3: The primal-dual algorithm has two names in literature: Primal-Dual Fixed-Point algorithm based on the Proximity Operator (PDFP$^2$O) and Proximal Alternating Predictor-Corrector (PAPC). 

4: In this paper, we extend it to solve $f(\vx)+h\square l(\vA\vx)$ with differentiable $l^*$ and prove its convergence under a weak condition (i.e., under a large dual stepsize). 

5: With additional assumptions, we show its linear convergence.

6: In addition, we show that this condition is optimal and can not be weaken.

7: This result recovers the recent proposed positive-indefinite linearized augmented Lagrangian method. 

8: 

9: Then we consider the application of this primal-dual algorithm in decentralized consensus optimization.

10: %Then we consider the dual problem of a decentralized consensus problem. 

11: We show that EXact firsT-ordeR Algorithm (EXTRA) and Proximal Gradient-EXTRA (PG-EXTRA) can be consider as the primal-dual algorithm applied on a problem in the form of $h\square l(A\vx)$. 

12: Then, the optimal upper bound of the stepsize for EXTRA/PG-EXTRA is derived. 

13: It is larger than the existing work on EXTRA/PG-EXTRA.

14: Furthermore, for the case with strongly convex functions, we proved linear convergence under the same condition for the stepsize.

15: %Therefore, we obtain an optimal upper bound of the stepsize for both convergence and linear convergence for EXTRA. 

16: 

17: 

18: \keywords{linearized augmented Lagrangian \and primal-dual \and decentralized consensus optimization}

19: % \PACS{PACS code1 \and PACS code2 \and more}

20: % \subclass{MSC code1 \and MSC code2 \and more}

21: \end{abstract}