1: \begin{abstract} 

2: This work describes a new variant of projective splitting for solving \myc{maximal} monotone 

3: inclusions \myc{and complicated convex optimization problems}. \myc{In the new

4: version}, cocoercive

5: operators can be processed with a single forward step per iteration. 

6: \myc{ In the convex optimization context, cocoercivity is equivalent to Lipschitz

7: differentiability.}

8: \myc{Prior forward-step versions of projective splitting did not fully exploit 

9: cocoercivity and required \emph{two} forward steps per iteration for such operators. 

10: Our new single-forward-step method} establishes a symmetry between projective

11: splitting algorithms, the classical forward-backward splitting method (FB),

12: and Tseng's forward-backward-forward method (FBF).

13: %Another symmetry is that 

14: The new procedure allows for larger stepsizes for cocoercive

15: operators: the stepsize bound is $2\beta$ for a $\beta$-cocoercive operator,

16: \myc{the same bound as has been established} for FB. We show that FB

17: corresponds to an unattainable boundary case of the parameters in the new

18: procedure. Unlike FB, the new method allows for a backtracking procedure when

19: the cocoercivity constant is unknown.

20: Proving convergence of the algorithm requires some departures from the \myc{prior}

21: proof framework for projective splitting.

22: We close with some computational tests establishing competitive performance 

23: for the method. 

24: \coapversion{\keywords{proximal operator splitting \and projective splitting \and convex nonsmooth optimization}}

25: \end{abstract}

26: