 |
Center for Nonlinear Studies |
Two key reasons for the success of mixed integer programming (MIP) are its modeling flexibility and the availability of extremely effective state-of-the-art solvers. Modeling problems with MIP can often be achieved with simple formulation techniques, but using more advanced techniques that are compatible with the solvers can significantly improve their performance. In this talk we use various examples to illustrate what can be modeled with linear and nonlinear MIP, and how using advanced techniques can provide a significant computational advantage. In particular, we show how formulation techniques can be used to transform a challenging simulation-based optimization problem arising in experimental design into a MIP that can be effectively tackled by state-of-the-art solvers.
Host: Russell Bent