On the Spectral Bias of Convolutional Neural Tangent and Gaussian Process Kernels

Part of Advances in Neural Information Processing Systems 35 (NeurIPS 2022) Main Conference Track

Bibtex Paper Supplemental


Amnon Geifman, Meirav Galun, David Jacobs, Basri Ronen


We study the properties of various over-parameterized convolutional neural architectures through their respective Gaussian Process and Neural Tangent kernels. We prove that, with normalized multi-channel input and ReLU activation, the eigenfunctions of these kernels with the uniform measure are formed by products of spherical harmonics, defined over the channels of the different pixels. We next use hierarchical factorizable kernels to bound their respective eigenvalues. We show that the eigenvalues decay polynomially, quantify the rate of decay, and derive measures that reflect the composition of hierarchical features in these networks. Our theory provides a concrete quantitative characterization of the role of locality and hierarchy in the inductive bias of over-parameterized convolutional network architectures.