1: \begin{abstract}

2: We propose a new first-order primal-dual optimization framework for a convex optimization template with broad applications. 

3: Our optimization algorithms feature optimal convergence guarantees under a variety of common structure assumptions on the problem template. 

4: Our analysis relies on a novel combination of three classic ideas applied to the primal-dual gap function: smoothing, acceleration, and homotopy. 

5: The algorithms due to the new approach achieve the best known convergence rate results, in particular when the template consists of only non-smooth functions.  

6: We also outline a restart strategy for the acceleration to significantly enhance the practical performance. 

7: We demonstate relations with the augmented Lagrangian method and show how to exploit the strongly convex objectives with rigorous convergence rate guarantees.

8: We provide numerical evidence with two examples and illustrate that the new methods can outperform the state-of-the-art, 

9: including  Chambolle-Pock, and the alternating direction method-of-multipliers algorithms.   

10: \vspace{1ex}

11: 

12: \noindent\textbf{Keywords:}

13: Gap reduction technique; first-order primal-dual methods; augmented Lagrangian; smoothing techniques; homotopy; separable convex minimization; parallel and distributed computation.

14: \end{abstract}

15: