Finite Neural Networks as Mixtures of Gaussian Processes: From Provable Error Bounds to Prior Selection

Open in new window