1: \begin{abstract}

2: Decentralized algorithms solve multi-agent problems over a connected network, where the information can only be exchanged with accessible neighbors. 

3: Though there exist several decentralized optimization algorithms, there are still gaps in convergence conditions and rates between decentralized algorithms and centralized ones. 

4: In this paper, we fill some gaps by considering two decentralized consensus algorithms: EXTRA and NIDS. 

5: Both algorithms converge linearly with strongly convex functions.

6: We will answer two questions regarding both algorithms. 

7: what are the optimal upper bounds for their stepsizes? Do decentralized algorithms require more properties on the functions for linear convergence than centralized ones?

8: More specifically, we relax the required conditions for linear convergence for both algorithms. 

9: For EXTRA, we show that the stepsize is in order of $O(\frac{1}{L})$ ($L$ is the Lipschitz constant of the gradient of the functions), which is comparable to that of centralized algorithms, though the upper bound is still smaller than that of centralized ones.

10: For NIDS, we show that the upper bound of the stepsize is the same as that of centralized ones, and it does not depend on the network. 

11: In addition, we relax the requirement for the functions and the mixing matrix, which reflects the topology of the network.

12: As far as we know, we provide the linear convergence results for both algorithms under the weakest conditions.

13: %and assume that the objective function is (restricted) strongly convex.

14: %A larger network-dependent stepsize is obtained to guarantee the linear convergence of EXTRA.

15: %For NIDS, the linear convergence still holds without harming the network independent stepsize.

16: \end{abstract}

17: