On the convexification of nonlinear programming problems: An applications-oriented survey (Q799590)

The mathematical programming problem P considered here consists in minimizing \(f_ 0(x)\) subject to \(f_ j(x)\leq 0\), \(j=1,...,m\) where \(f_ j(x)\), \(j=0,...,m\) are real-valued functions defined on a convex set \(C\subset {\mathbb{R}}^ n\). The object of the paper is to study the transformability of P into a convex programming problem. The paper consists of five sections. Section One is introductory. Section Two deals with the functions which can be transformed into convex functions by means of strictly increasing functions (range transformation). It follows from the results obtained that it is often sufficient to prove the existence of such a transformation for each of the functions \(f_ j(x)\), 0\(\leq j\leq m\). Section Three is devoted to convexification by domain transformation (variable transformation). Two main examples of problems which can be convexified in this way are the convex-concave fractional programming problem and the classical geometric programming problem. Section Four ''Objective function approximation'' deals with problems which have convex constraints and a non-convex objective function which cannot be convexified. The so-called natural approximation and the use of convex envelopes are discussed. The last section presents a summary and some suggestions for further research.