Codes for quasi-Monte Carlo feature maps for Gaussian kernels in MATLAB.
Given any pair of points x and y, compute a feature map z(x) and z(y) such that k(x,y) is approximated by the inner product between z(x) and z(y) where in this case k is the Gaussian kernel.
See qmc_faeture_maps_complex.m
for usage.
See demo.m
for a simple example.
J. Yang, V. Sindhwani, H. Avron, and M. W. Mahoney, Quasi-Monte Carlo Feature Maps for Shift-Invariant Kernels. International Conference on Machine Learning (ICML), 2014.