I took an optimization methods course at UW CFRM and solver was mentioned as a way to set up toy models in a pinch, but in general the course covered more polished LP/QP and DE solvers that would be used in an institutional framework, in addition to the underlying methods of various solver algorithms. I don't recall any UW course seriously using Excel's Solver.
As the instructor for UW CFRM 555 Optimization Methods in Finance course, I am providing the following information about the course and our use of the Solver in Excel.
The Solver bundled with Excel is a very useful tool, but within the context of Excel can only address models of a limited size. The familiarity that most students have with spreadsheet based computations makes it a very useful starting place for introduction of optimization software, but it is not typically appropriate for full scale models. The Excel Solver is "taught" in the sense that we work through example problems and review the sensitivity analysis and other optional reports that are available within the tool. I generally find that many students already have some experience using the solver, but few if any are familiar with the range of available output.
The CFRM555 course introduces students to other optimization software including (among others) the glpk and quadprog packages in R (a platform with which most UW students are familiar) and commercial solvers accessible through a course license to CPLEX provided by the IBM Academic Initiative and the AMPL modeling language provided by the AMPL for Courses program. The AMPL language provides a convenient interface to a number of extremely robust solvers either bundled with AMPL or available through NEOS (NEOS Server: State-of-the-Art Solvers for Numerical Optimization at
www.neos-server.org). Based on submissions to the NEOS website, AMPL is the most common tool for specifying optimization models and submitting them for solution.
There are a number of other fine products that could be incorporated into a course such as CFRM555 including Matlab, Mathematica and TK Solver, but within the limits of the course setting it is difficult to cover all the possibilities.
Primary goals of the course include not just familiarity with current optimization tools, but more critically the ability to identify real world problems as appropriate applications for optimization techniques and the ability to translate these real world situations into mathematical models to which appropriate solution software can be applied. CFRM555 includes exposure to the theory behind optimization approaches including an in depth review of duality, definitions of efficiency for algorithms, use of linear approximations and decomposition as well as a review of heuristic approaches for problems that don't lend themselves to more theoretically elegant solutions.
--Steve Murray
Steven Murray | Computational Finance & Risk Management