1: \begin{abstract}

2: \makeatletter{}This paper considers a conceptual version of a convex optimization algorithm

3: which is based on replacing a convex optimization problem

4: with the root-finding problem for the approximate sub-differential mapping

5: which is solved by repeated projection onto the epigraph of conjugate function.

6: Whilst the projection problem is not exactly solvable in finite space-time

7: it can be approximately solved up to arbitrary precision by simple iterative

8: methods, which use linear support functions of the epigraph.

9: It seems therefore useful to study computational characteristics of the idealized

10: version of this algorithm when projection on the epigraph is computed precisely to estimate the potential benefits

11: for such development.

12: The key results of this study are that the conceptual algorithm

13: attains super-linear rate of convergence in general convex case,

14: the rate of convergence becomes quadratic for objective functions

15: forming super-set of strongly convex functions,

16: and convergence is finite when objective function has sharp minimum.

17: In all cases convergence is global and does not require differentiability of the objective.

18:  

19: {\bf Keywords:} convex optimization, conjugate function, approximate sub-differential,

20: super-linear convergence, quadratic convergence, finite convergence, projection, epigraph

21:  

22: \end{abstract}

23: