Why build a reusable neighbor graph when comparing several downstream graph-based estimators on the same representation?
Classic MLLevel: IntermediateNeighbor graph caching
Why build a reusable neighbor graph when comparing several downstream graph-based estimators on the same representation?
Hints
- Separate graph construction from graph consumption.
- Reuse is valid only for compatible inputs and settings.
Answer
Neighbor search can be a major shared cost. Reusing a compatible graph avoids repeated searches, but graph parameters and training/validation boundaries must still match each experiment.
Source and licence: scikit-learn 1.7.2 — neighbors · BSD-3-Clause · answer: Quantoros editorial