Training Data
Training data is the set of examples used to teach a machine learning model, from which it learns its parameters.
What is Training Data?
The EU AI Act defines training data as data used to fit an AI system's learnable parameters. Its quality, relevance, representativeness and lawful sourcing largely determine how well and how fairly a model performs. Copyright, personal data and bias in training data are major legal issues for AI developers.
How it works
Data is collected, cleaned, labelled where needed and fed to the learning algorithm, usually many times over, while separate validation and test sets are held back to check performance.
← All Artificial Intelligence & Superintelligence definitions · ← Back to the A–Z definitions