> These somehow use both X (the amplitude) and t (time) in devising an error metric.

Yes this error metric need not be correlation nor Euclidean