4.1. Synthetic testing of cDTW

Various tests were performed for the DTW algorithm. In general, the performance on simple depth shifting patterns is good and robust.

– Care must be taken with low-frequency baseline shifts: These can strongly influence the result depending on the metrics (e.g. euclidean / least squares) chosen.

– Typically, with traditional local constraints, the DTW output mapping can simulate sedimentation rate changes through a pulse-code modulation (rapid on/off patterns) that require smooting for a more realistic sedimentation rate / mapping (see below on right). This is particularly the case if a specific local pattern is specified (e.g. Itakura, 1975, with possible sed. rates of 0.5, 1.0, 2.0)

Fig. 4.2 General testing of performance of cDTW Tool(from [G2])

Fig. 4.3 (cont). General testing of performance of cDTW Tool (from [G2])

An example for bandpass constraints is shown below:

Fig. 4.4 (cont). Testing of performance of cDTW Tool (from [G2]) with band-pass constraints.

A test of various relaxation settings on real-data vs. astronomical target curves:

Fig. 4.5 (cont). Testing of performance of cDTW Tool (from [G2]) on real geological data.