arXiv.itsnick.co

an independent field guide to 23 million papers from arXiv & OpenAlex — not affiliated with arXiv.org

Computer Science · Artificial Intelligence

ACCEPTED BY IEEE TNNLS 1 Linear Regression Based Efficient SVM Learning for Large Scale Classification

Jianxin Wu, Hao Yang

See on the map

Abstract

Abstract—For large scale classification tasks, especially in the classification of images, additive kernels have shown state-of-the-art accuracy. However, even with the recent development of fast algorithms, learning speed and the ability to handle large scale tasks are still open problems. This paper proposes algorithms for large scale SVM classification and other tasks using additive kernels. First, a Linear Regression SVM (LR-SVM) framework for general non-linear kernel is proposed, by using linear regression to approximate gradient computations in the learning process. Second, we propose a Power Mean SVM (PmSVM) algorithm for all additive kernels, by using non-symmetric explanatory variable functions. This non-symmetric kernel approximation has advantages over existing methods: it does not require closed-form Fourier transforms, and it does not require extra training for the approximation either. Compared on benchmark large scale classification datasets with millions of examples or millions of dense feature dimensions, PmSVM has achieved the highest learning speed and highest accuracy among recent algorithms in most cases. Index Terms—Large scale classification, additive kernels, linear regression, SVM, Nyström approximation. I.

More like this