Quantifying the Benefit of Using Differentiable Learning over Tangent Kernels

Open in new window