About the project

Understanding comes first.

TensorAtlas is an open atlas of machine learning concepts, designed to make room for the ideas themselves.

A small beginning

This first edition connects 22 short introductions across mathematics, learning methods, architectures, evaluation, and deployment. Follow prerequisites, explore a branch, or browse the list.

Made to be explored

Each introduction pairs an intuitive explanation with a concrete example. Most include a self-check; gradient descent includes a static diagram. Interactive experiments are next. Pages are designed for a phone as carefully as a larger screen.

Built in the open

Ideas, corrections, and contributions are welcome on GitHub.