A case for new neural network smoothness constraints

Open in new window