Successive Overrelaxation for Support Vector Regression

Support vector regression (SVR) is an important tool for data mining. In this paper, we first introduce a new way to make SVR have the similar mathematic form as that of support vector classification. Then we propose a versatile iterative method, successive overrelaxation, for the solution of extremely large regression problems using support vector machines. Experiments prove that this new method converges considerably faster than other methods that require the presence of a substantial amount of the data in memory.