Parallel algorithm for support vector machines training and quadratic optimization problems

Abstract

We consider an iterative algorithm, suitable for parallel implementation, to solve convex quadratic optimization problems with a single constraint and simple bounds on the variables. This approach can be used to effectively solve the quadratic optimization problem arising in training support vector machines. The proposed algorithm is a double iterative schema based on inexact matrix splitting and alternating direction method.

Keywords:

• Support vector machines
• Quadratic programming
• Matrix splitting methods
• Alternating direction methods