Recording the origin of each data source and every: Which activity most directly fulfills that requirement?

Lineage and provenance capture the origin of data and the transformations it undergoes on the path into the training set.

The question

Preparing for AIGP? Take the free 5-min readiness quiz →

1. Recording the origin of each data source and every transformation applied on the way to the training set. ✓

   Lineage and provenance describe where data came from and how it was transformed and moved into the training set.
2. Measuring the trained model's predictive accuracy against a held-out validation set before its public release.

   Plausible but this is validation testing of model performance, not documentation of the data's origin and path.
3. Cataloging the compute resources and the specific hyperparameters that were consumed during each training run.

   Plausible but this records the training configuration, not the source and transformation history of the data.
4. Listing the access permissions and the encryption keys applied to the stored training dataset.

   Plausible but this is data security control, not a record of where the data originated and how it changed.

How to remember it

Lineage and provenance capture the origin of data and the transformations it undergoes on the path into the training set.

More Understanding How to Govern AI Development questions

• It is not fit-for-purpose despite lawful rights and high: Under data-governance requirements, what is the most →
• Examining whether model outcomes differ unjustifiably: Which purpose is served specifically by the bias →
• It should be logged and managed as an identified risk: As a matter of good AI development governance, what →
• All 426 Understanding How to Govern AI Development questions →