On a box-constrained linear symmetric cone optimization problem (Q2420823): Difference between revisions

The authors analyze the box-constrained linear symmetric cone optimization problem written as \(\min \left\langle c_{1},x_{1}\right\rangle +\sum_{i=1}^{r}\left\langle c_{2,i},x_{2,i}\right\rangle +\sum_{i=1}^{t}\left\langle c_{3,i},x_{3,i}\right\rangle \) under the constraints \ \(0\preceq_{1}x_{1}\preceq_{1}b_{1}\), \(0\preceq_{2}x_{2,i}\preceq_{2}b_{2,i}\), \(0\preceq_{3}x_{3,i}\preceq_{3}b_{3,i}\). Here \(\preceq_{1}\) refers to \(\mathbb{R}_{m}^{+}\), \(\preceq_{2}\) refers to a \(l\)-dimensional second-order cone and \(\preceq_{3}\) refers to a \(n\) -dimensional semidefinite matrix cone. The authors first recall some properties of symmetric cones. They then analyze the separate minimization of each of the three quantities which appear in the above problem. For each minimization problem, they give the expression of the minimum objective value and of the optimal solution. In the two last cases, they assume that \( 0\prec_{2}b_{2}\) or \(0\prec_{3}b_{3}\). They finally give the expression of the minimum objective value and of the optimal solution set of the above problem, assuming that \(0\prec b\). They then consider four application cases: a linear symmetric cone optimization problem with two symmetric cone constraints, a linear second-order cone optimization problem with a linear constraint, a nonlinear symmetric cone optimization problem with a box-constraint and a linear symmetric cone optimization problem. In the third and fourth cases, the authors also propose an algorithm either based on direct arguments or on a BFGS quasi-Newton method. The paper ends with the presentation of numerical results obtained with these two algorithms.