Autograd Release

Autograd, a predecessor to JAX, was released, providing automatic differentiation for NumPy code. It showcased the power of flexible, Pythonic autodiff.

Significance

Autograd demonstrated the feasibility and appeal of automatic differentiation for NumPy, laying the conceptual and practical groundwork for JAX's core `grad` functionality and flexible design.

Key facts

Date
2015-01-01
Type
invention
Location
Harvard University / Google