1: \begin{abstract}

2: This paper provides a theoretical and numerical comparison of 

3: classical {first-order} splitting methods for solving smooth convex 

4: optimization problems and cocoercive equations. {From}  a 

5: theoretical point 

6: of view, we compare convergence rates of gradient descent, 

7: forward-backward, Peaceman-Rachford, and Douglas-Rachford 

8: algorithms for minimizing the sum of two smooth convex functions 

9: when one of them is strongly convex. A similar comparison is given in 

10: the more general cocoercive setting under the presence of strong 

11: monotonicity and we observe that the convergence rates in 

12: optimization are strictly better than the corresponding rates for 

13: cocoercive equations for some algorithms.

14: We obtain improved rates with respect to the literature in several 

15: instances {by}  exploiting the structure of our problems. 

16: Moreover, we  

17: indicate { which algorithm has the lowest convergence rate 

18: depending on 

19: strong 

20: convexity and cocoercive parameters.}

21: {From}  a numerical point of view, we verify our theoretical results 

22: by implementing and comparing previous algorithms in well 

23: established signal and image inverse problems involving sparsity. 

24: We replace the widely used $\ell_1$ norm {with}  the Huber loss 

25: and we observe that 

26: fully proximal-based strategies have numerical and theoretical 

27: advantages with respect to methods using gradient steps. 

28: \footnote{This work is supported by Agencia Nacional de 

29: Investigaci\'on 

30: y Desarrollo (ANID) from Chile, under grant FONDECYT 1190871, by 

31: Centro de

32: Modelamiento Matem\'atico (CMM), ACE210010 and FB210005, 

33: BASAL funds for centers of excellence, and also by the ANR (Agence 

34: Nationale de la Recherche) from France

35: ANR-19-CE48-0009 Multisc'In. The authors acknowledge the support 

36: from 

37: LIA-MSD, CNRS-France.}

38: \end{abstract}

39: