A Conceptual Conjugate Epi-Projection Algorithm of Convex Optimization: Superlinear, Quadratic and Finite Convergence

TL;DR

This paper introduces a conceptual convex optimization algorithm based on conjugate epiprojection, demonstrating superlinear, quadratic, and finite convergence rates under various conditions, without requiring differentiability.

Contribution

It proposes a novel conceptual algorithm using conjugate epiprojection with theoretical convergence guarantees in convex optimization.

Findings

Achieves super-linear convergence in general convex problems.

Attains quadratic convergence for strongly convex functions.

Ensures finite convergence for functions with sharp minima.

Abstract

This paper considers a conceptual version of a convex optimization algorithm whic is based on replacing a convex optimization problem with the root-finding problem for the approximate sub-differential mapping which is solved by repeated projection onto the epigraph of conjugate function. Whilst the projection problem is not exactly solvable in finite space-time it can be approximately solved up to arbitrary precision by simple iterative methods, which use linear support functions of the epigraph. It seems therefore useful to study computational characteristics of the idealized version of this algorithm when projection on the epigraph is computed precisely to estimate the potential benefits for such development. The key results of this study are that the conceptual algorithm attains super-linear rate of convergence in general convex case, the rate of convergence becomes quadratic for objective functions forming super-set of strongly convex functions, and convergence is finite when objective function has sharp minimum. In all cases convergence is global and does not require differentiability of the objective. Keywords: convex optimization, conjugate function, approximate sub-differential, super-linear convergence, quadratic convergence, finite convergence, projection, epigraph