This repo contains a set of Jupyter notebook describing how to use various autograd functionalities, complementing the excellent tutorial located at the repo itself, including:
-
basic_autograd_examples.ipynb covering basic functionalities such as: derivative computation using standard and lambda functions, subtleties involved in automatic differentiation and the array of gradient prototypes provided by
autograd, and computing partial derivatives of multi-input functions -
flattening_functions_using_autograd.ipynb covering usage of
autograd's flatten_func function
These notebooks were produced as supplementary material for the second edition of the textbook Machine Learning Refined, published Cambridge University Press, set for release in mid-2019. You can find a host of examples employing autograd and - in particular - flatten_func on the main repository for the textbook located here (see for example the drafts on multi-class classification fully connected networks).