Scaling up Kernel SVM on Limited Resources: A Low-rank Linearization Approach

Kernel Support Vector Machine delivers state-of-the-art results in non-linear classification, but the need to maintain a large number of support vectors poses a challenge in large scale training and testing. In contrast, linear SVM is much more scalable even on limited computing recourses (e.g. daily life PCs), but the learned model cannot capture non-linear concepts. To scale up kernel SVM on limited resources, we propose a low-rank linearization approach that transforms a non-linear SVM to a linear one via a novel, approximate empirical kernel map computed from efficient low-rank approximation of kernel matrices. We call it LLSVM (Low-rank Linearized SVM). We theoretically study the gap between the solutions of the optimal and approximate kernel map, which is used in turn to provide important guidance on the sampling based kernel approximations. Our algorithm inherits high efficiency of linear SVMs and rich repesentability of kernel classifiers. Evaluation against large-scale linear and kernel SVMs on several truly large data sets shows that the proposed method achieves a better tradeoff between scalability and model representability.