The Solving Convex Quadratic Programming Problems Arising

in Support Vector Machine Framework

ABSTRACT

Support vector machines are a collection of techniques to solve a classification problem of a regression problem. In their core lies a convex quadratic optimization problem (QP). This convex QP is often very large scale and the Hessian of the objective function is completely dense. Hence, using conventional optimization techniques may be too expensive. However, these problems may have some inherent structure, which we can use to create efficient algorithms. We will discuss two main approaches to solving these problems: one based on interior point methods and one based on active set methods.