Last year we held a machine learning seminar in our London office, which was an opportunity to reproduce some classical deep learning results with a nice twist: we used OCaml as a programming language rather than Python. This allowed us to train models defined in a functional way in OCaml on a GPU using TensorFlow.
