NeurIPS 2020

Sparse Spectrum Warped Input Measures for Nonstationary Kernel Learning


Meta Review

Three reviewers support acceptance, and one reviewer supports rejection. R1 rejects on the basis that the idea of using input-warping with sparse Gaussian processes is straightforward, so does not represent a novel contribution. The rebuttal rejoins that the simplicity of the approach is a boon to the practitioners: I concur with this view: the combination is straightforward, but the empirical results show that it is empirically useful. R2, R3, and R4 support the paper in light of the fact that it provides a method for modeling non-stationary with Gaussian processes that exhibits good empirical performance. I concur with reviewers R2, R3, and R4, and accept this paper.