Warm Start
Noun · AI & Machine Learning
Definitions
Starting training or optimization from an existing model, checkpoint, or good prior state instead of beginning from random initialization. Warm starts are often used to save time and preserve useful prior learning.
In plain English: Starting from an existing good state instead of from scratch.
Example: "They used a warm start from the earlier support model so the new run converged much faster on the updated policy data."