It has been a while that I am trying to understand the dtw (dynamic time warping) algorithm but I steel can't figure out the loginc behind this algorithm..
In many references it's explained that to compute the path between to time series I need to implement this recursivity:
I have many questions here and I hope some one can give me an answer:
- What we mean exactly by the path between two time series ?
- What is the (cost path matrix described here:
- Why not just compute the distances between a given point of
t
and other points ofr
and select the closest one an repeat this until matching all points fromt
to points fromr
??
Thanks in advance for any helpfull answer / comment.