Efficient Training of Infinite-depth Neural Networks via Jacobian-free Backpropagation A promising trend in deep learning replaces fixed depth models by approximations of the limit as network depth ...