On discretisation drift and smoothness regularisation in neural network training

Open in new window