Feature store is a data management layer for machine learning with an aim to share & discover features and increase efficiency for machine learning (it is typically extended part for MLOps solutions). Use this tag for questions about the data management layer for machine learning (feature-store concept covers API, persistent layer, UI, etc.) on various platforms.
Feature store is a data management layer for machine learning with an aim to share & discover features and increase efficiency for machine learning (it is typically extended part for MLOps solutions). The Feature store has these three key parts, the meta repository, the on-line part and the off-line part.
The meta repository keeps definition of ML/AL models (view to feature sets, feature vectors, features, entities, etc.) The on-line part supports real-time responses (with support event and API centric views) and ML/AI executions and the off-line part supports procession of batch/big data (data centric view, e.g. an aggregation of long data, ML/AI experiments, etc.).
The Feature store extents MLOps solutions and support the mesh data concept.
More information
- Feature Store, common view
- What is Feature Store
