Sparse Kernel SVMs via Cutting-Plane Training

We explore an algorithm for training SVMs with Kernels that can represent the learned rule using arbitrary basis vectors, not just the support vectors (SVs) from the training set. This results in two