Exploiting structure in outer approximation based approaches for SOCPAlgebra & Discrete Mathematics
|Sarah Drewes, TU Darmstadt / UC Berkeley
|Thu, Apr 7 2011, 4:10PM
We discuss outer approximation based approaches to solve mixed integer second order cone programming problems. These methods utilize a linear outer approximation to derive integer assignments that are then used to solve the corresponding continuous second order cone problems. The derived solution is again used to strengthen the approximation. Results for the case of mixed integer nonlinear programs with continuously differentiable constraint functions can be extended to guarantee convergence of this approach under similar assumptions. We show how this method can be sped up significantly by exploiting the symmetric structure of a particular class of mixed 0-1 second order cone problems. A computational study based on application problems is provided.