mygrad 2.2.0

Brings drop-in automatic differentiation to NumPy

MyGrad is a lightweight library that adds automatic differentiation to NumPy – its only dependency is NumPy! It’s primary goal is to make automatic differentiation an accessible and easy to use across the Python/NumPy ecosystem.

MyGrad introduces a tensor object, which behaves like NumPy’s ndarray object, but that builds a computational graph, which enables MyGrad to perform reverse-mode differentiation (i.e. “backpropagation”). By exploiting NumPy’s mechanisms for ufunc/function overrides, MyGrad’s tensor works “natively” with NumPy’s suite of mathematical functions so that they can be chained together into a differentiable computational graph.

NumPy’s systems for broadcasting operations, producing views of arrays, performing in-place operations, and permitting both “basic” and “advanced” indexing of arrays are all supported by MyGrad to a high-fidelity.