Computer Science · Computer Vision and Pattern Recognition
Hash-SVM: Scalable Kernel Machines for Large-Scale Visual Classification
See on the mapAbstract
This paper presents a novel algorithm which uses hash bits for efficiently optimizing non-linear kernel SVM in very large scale visual classification problems. Our key idea is to represent each sample with compact hash bits and define an inner product over these bits, which serves as the sur-rogate of the original nonlinear kernels. Then the problem of solving the nonlinear SVM can be transformed into solv-ing a linear SVM over the hash bits. The proposed Hash-SVM enjoys both greatly reduced data storage owing to the compact binary representation, as well as the (sub-)linear training complexity via linear SVM. As a crucial component of Hash-SVM, we propose a novel hashing scheme for arbi-trary non-linear kernels via random subspace projection in reproducing kernel Hilbert space. Our comprehensive anal-ysis reveals a well behaved theoretic bound of the deviation between the proposed hashing-based kernel approximation and the original kernel function. We also derived moderate requirements on the hash bits for achieving a satisfactory accuracy level. Several experiments on large-scale visu-al classification benchmarks are conducted, including one with over 1 million images. The results well demonstrated the superiority of our algorithm when compared with other alternatives. 1.