Tutorials

The tutorials show how XDFlow uses metadata as pipeline context: coordinate targets, named-dimension transforms, leakage-safe validation, fold-invariant reuse, modular pipelines, and specialized split policies.

• 5-Minute Core Quickstart: runnable base-install pipeline with coordinate targets, stratified CV, stateful refits, and prediction alignment
• Hyperparameter Tuning: Optuna-backed search over pipeline parameters, optional steps, switch choices, and multiple pipelines
• Spectral Pipeline Walkthrough: signal-processing pipeline where expensive fold-invariant feature extraction can be reused across folds
• Reusable ML Patterns: multilabel prediction, class/domain weighting, and few-shot domain transfer without side-channel label or split bookkeeping

The original notebook and dataset helper are still kept under docs/tutorials/ for local exploration, but the Markdown walkthroughs are the canonical versions for the published docs site.