A self consistent theory of Gaussian Processes captures feature learning effects in finite CNNs

Open in new window