Bernstein-Schur Kernels: Random Features by Sketched Modulation and Radial Randomization

In three linesBernstein-Schur kernels: random feature construction combining sketched finite modulation and randomized completely monotone radial factors. Feature dimension = Dm (sketch size m × radial draws D), avoiding O(d²) dependence. Guarantees: unbiasedness, operator-norm bounds controlled by intrinsic dimensions, spectral stability for kernel ridge regression.

Summary generated by Claude — human-verified