Efficient χ2 Kernel Linearization via Random Feature Maps

doi:10.1109/TNNLS.2015.2476659

Efficient χ2 Kernel Linearization via Random Feature Maps

2016-11-01

6

PRE

AI

DOI：10.1109/TNNLS.2015.2476659

Original

Original request for help

Save

Abstract

En 中文

Explicit feature mapping is an appealing way to linearize additive kernels, such as chi(2) kernel for training large-scale support vector machines (SVMs). Although accurate in approximation, feature mapping could pose computational challenges in high-dimensional settings as it expands the original features to a higher dimensional space. To handle this issue in the context of chi(2) kernel SVMs learning, we introduce a simple yet efficient method to approximately linearize chi(2) kernel through random feature maps. The main idea is to use sparse random projection to reduce the dimensionality of feature maps while preserving their approximation capability to the original kernel. We provide approximation error bound for the proposed method. Furthermore, we extend our method to chi(2) multiple kernel SVMs learning. Extensive experiments on large-scale image classification tasks confirm that the proposed approach is able to significantly speed up the training process of the chi(2) kernel SVMs at almost no cost of testing accuracy.

Keywords:

chi(2) kernel

feature maps

image classification

sparse random projection

support vector machines (SVMs)

Efficient χ2 Kernel Linearization via Random Feature Maps

Abstract

AI Summary

Journal

Organization