paper-conference

We develop an Gaussian Process fourier features approximation using a custom quadrature rule targeted at the trignometric form of the kernel function’s spectral form. Error analysis shows superior approximation performance relative to Random and Gaussian quadrature fourier feature method.