Training technique that supplements data by finding and using similar examples from a database to improve model generalization.