Transferring localized facial learning across all of face space