Canonical d. c. programming techniques for solving a convex program with an additional constraint of multiplicative type (Q685859)

Canonical d.c. programming techniques were developed for solving optimization problems in which every function is represented as a difference of two convex functions. The author presents an application of canonical d.c. programming techniques to solve a nonconvex programming problem of minimizing a linear function \(cx\) over a convex set \(X\subset\mathbb R^ n\) with an additional constraint of multiplicative type \(\prod^ p_{i=1} \psi_ i(x)\leq 1\), where the functions \(\psi_ i\) are convex and positive on \(X\). The approach is based on transforming the original optimization problem, by using \(p\) additional variables, into a canonical d.c. programming problem with the special structure that the reverse convex constraint involved does only depend on the newly introduced variables. This special structure suggests a few modifications of certain techniques in d.c. programming in a way that the operations handling the nonconvexity are actually performed in the space of the additional variables. The convergence of the algorithm is proved and some computational results are also provided.