Sphere Support Vector Machines for large classification tasks

摘要: This paper introduces Sphere Support Vector Machines (SVMs) as the new fast classification algorithm based on combining a minimal enclosing ball approach, state of art nearest point problem solvers and probabilistic techniques. The blending three significantly speeds up training phase SVMs also attains practically same accuracy other models over several large real datasets within strict validation frame double (nested) cross-validation. results shown are promoting SphereSVM outstanding alternatives for handling ultra-large in reasonable time without switching to various parallelization schemes SVM algorithms recently proposed.