Learnability of convolutional neural networks for infinite dimensional input via mixed and anisotropic smoothness.

Sho Okumoto, Taiji Suzuki

Browse the full ICLR paper archive.