Skip to content

Start

Start here

Three pages: installation, a complete worked example, and the concepts the rest of the documentation builds on. They are the only prerequisites for the other sections.

  • Install

    Installation with pip or uv, the optional HPO extra, and notes on GPU support.

  • First forecast

    A complete worked example: train an Echo State Network on a chaotic attractor, then forecast it autoregressively.

  • The mental model

    The four ideas the whole library is built on: frozen dynamics, the state, the algebraic readout, and the two-phase forecast.