I would like to know the method to solve quadratic programming with computer (not by hand). Any mathematical material I can re开发者_高级运维ad? Thanks!
You want to use a black-box software package. It is notoriously difficult to implement a robust QP solver.
You didn't mention what language you are using, but if it's C++, QuadProg++ is light-weight and free.
There can be used some kind of interior point methods. Often QP problem is very special and it is possible to use less general methods, for example Platt's algorithm for QP arising in SVM.
If the code is short and uses SVD and LU, then it's probably a simple solver for problems with only equality constraints or no constraints at all. QP with inequality constraints can be solved using tools such as quadprog(matlab), quadprog++(c++), mosek(matlab and c++). c++ libraries can be used in .net language.
The answer strongly depends on the size of your problem, and the convexity of the quadratic functions. If your problems are convex and you like using python, you can use cvxmod for free. For mathematical material, there's also the convex optimization book, freely available. For non-convex problems, the package IPopt from the coin-or project will work to find a locally-optimal solutions.
精彩评论