Data dependent kernels in nearly-linear time
Guy Lever, Tom Diethe, John Shawe-Taylor ; JMLR W&CP 22: 685-693, 2012.
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.