A branch and bound algorithm for solving a class of D-C programming (Q1780534)
From MaRDI portal
scientific article
| Language | Label | Description | Also known as |
|---|---|---|---|
| English | A branch and bound algorithm for solving a class of D-C programming |
scientific article |
Statements
A branch and bound algorithm for solving a class of D-C programming (English)
0 references
13 June 2005
0 references
A special class of DC programming problems is considered where the objective function is the sum of a convex quadratic function, \(p(x)\), and a separable concave function, \(\varphi(x)= \varphi_1(x_1)+\cdots+ \varphi_n(x_n)\). The feasible domain is the intersection of a polytope and a box. The problem is converted into a quadratic convex problem by replacing the functions \(\varphi_j(x_j)\) by linear underestimates \(\psi_j(x_j)\), which depend on the subdomains generated by the branch and bound algorithm. Several bisection techniques are compared in numerical examples.
0 references
DC programming
0 references
branch and bound algorithm
0 references
\(\omega\)-subdivision
0 references
largest distance bisection
0 references
normal rectangular subdivision
0 references
numerical examples
0 references
0 references
