An automatic differentiation library for my CS207 final project at Harvard University.
autodiff is a package that provides users with forward-mode and reverse-mode automatic differentiation tools to compute derivatives of mathematical functions up to machine precision with high efficiency. With this software, the user can perform automatic differentiation for themselves. Our software focuses on the elementary functions of automatic differentiation for users who are more unfamiliar with the concept to utilize.
The library currently supports forward mode and reverse mode automatic differentiation, real functions with scalar or vector inputs, and multiple functions with scalar or vector inputs.
Forward mode is implemented using the unique properties of dual numbers. Reverse mode is implemented using reverse traversal through a computational graph.
The user can install the package using the code below
pip install --index-url https://test.pypi.org/simple/ --extra-index-url https://pypi.org/simple AutoDiff-Library==1.0
Please take a look at the documentation under the "How to Use autodiff" section.
Code coverage results can be found here.
Package on Test PyPI can be found here.
