AI Glossary · Data & Platform
Training Data
The corpus a model learns from; governance covers provenance, licensing, PII, consent and representativeness.
Definition
What is Training Data?
Training Data is the corpus a model learns from; governance covers provenance, licensing, PII, consent and representativeness.
- Category
- Data & Platform
- Glossary set
- 14 related terms
- Audience
- Enterprise AI leaders
Why does Training Data matter in enterprise AI?
Training Data matters in enterprise AI platforms because it affects data readiness, governance, scalability, and the operating foundation for AI use cases.
Related terms in Data & Platform
- AUC (Area Under Curve)
- A single-number summary of a classifier's ability to separate classes across all thresholds; useful for imbalanced problems like fraud or churn.
- Bring Your Own Model (BYOM)
- A platform pattern that lets enterprises deploy fine-tuned or proprietary models inside a vendor's runtime instead of using only the vendor's default model.
- Data Fabric
- An architecture that unifies distributed data via metadata, semantic models and active governance — the layer AI depends on to reason across silos.
- Data Lakehouse
- A unified store combining lake economics with warehouse governance (Databricks, Snowflake, Iceberg). The default AI training and analytics substrate.
- Embedding
- A numeric vector representing meaning of text, image or audio, enabling semantic search, clustering and retrieval-augmented generation.
- Feature Store
- A managed catalog of reusable ML features with online and offline serving — key to consistent training and inference.