Fundamentals

If you are new to GraphReduce, start here before diving into end-to-end examples.

GraphReduce turns relational feature engineering into a graph execution problem:

1. Define tables/files as nodes.
2. Define foreign-key relationships as edges.
3. Apply deterministic operations at each node.
4. Roll child features up to parent grain.
5. Join rolled features back to parent nodes.
6. Optionally compute labels at the target grain.

Core pages:

• Operations pipeline
• Joins and cardinality
• Time windows and leakage safety

Then continue with:

• Abstractions
• Quickstart tutorial
• End-to-end examples