Data-dependent kernels in nearly-linear time
Machine Learning
2015-03-19 v1
Abstract
We propose a method to efficiently construct data-dependent kernels which can make use of large quantities of (unlabeled) data. Our construction makes an approximation in the standard construction of semi-supervised kernels in Sindhwani et al. 2005. In typical cases these kernels can be computed in nearly-linear time (in the amount of data), improving on the cubic time of the standard construction, enabling large scale semi-supervised learning in a variety of contexts. The methods are validated on semi-supervised and unsupervised problems on data sets containing upto 64,000 sample points.
Cite
@article{arxiv.1110.4416,
title = {Data-dependent kernels in nearly-linear time},
author = {Guy Lever and Tom Diethe and John Shawe-Taylor},
journal= {arXiv preprint arXiv:1110.4416},
year = {2015}
}