Convexoid: A Minimal Theory of Conjugate Convexity

TL;DR

Convexoid introduces a minimal theoretical framework for conjugate convexity, generalizing classical duality results and supporting diverse approximation schemes and structures in convex optimization.

Contribution

It establishes a minimal system called convexoid that derives key convex analysis concepts and duality results, broadening the scope of convex optimization theory.

Findings

Derived conjugate functions and subdifferentials within convexoid

Generalized duality conditions including weak and strong duality

Supported various approximation schemes and structural representations

Abstract

A key idea in convex optimization theory is to use well-structured affine functions to approximate general functions, leading to impactful developments in conjugate functions and convex duality theory. This raises the question: what are the minimal requirements to establish these results? This paper aims to address this inquiry through a carefully crafted system called the convexoid. We demonstrate that fundamental constructs, such as conjugate functions and subdifferentials, along with their relationships, can be derived within this minimal system. Building on this, we define the associated duality systems and develop conditions for weak and strong duality, generalizing the classic results from conjugate duality and radial duality theories. Due to its flexibility, our framework supports various approximation schemes, including approximating general functions using symmetric-conic, bilinear, radial, or piecewise constant functions, and representing general structures such as graphs, set systems, fuzzy sets, or toposes using special membership functions. The associated duality results for these systems also open new opportunities for establishing bounds on objective values and verifying structural properties.