The paper referenced here is well beyond “just deep learning”.