A Proximity Algorithm for Support Vector Machine Classification
Antonios V. Sideris, S.E. Castella · 2006
We propose a new algorithm for Support Vector Machine classification based on a geometric interpretation of the problem as finding the minimum distance between the polytopes defined by the points of the two classes. This geometric formulation applies to the hard margin, and the soft margin classification problem with quadratic violations. Our approach is based on Wolfe's classical proximity algorithm and our results show that the computational and storage requirements per iteration are relatively modest.