Quadratic Programming based on algorithms from computational geometry Vojtech Franc We address a problem of a large-scale Quadratic Programming (QP) optimization. This type of optimization problems occurs, e.g., in machine learning or control theory. Conventional optimization methods become infeasible for problems with the number of variables exceeding several thousands. We showed that special instances of the large-scale QP tasks can be solved using simple algorithms from computational geometry. The idea will be demonstrated on the Generalized Minimal Norm Problem (GMNP) which is an example of such QP tasks. We describe several simple yet efficient algorithms to solve the GMNP. Finally, we give a sample of applications including classifier learning and density estimation.