WHY longer/shorter lines?

[xiaoFine]

Hi apolcyn! Awesome implementation of TCMM!
BTW, I wonder why you need to determine the longer/shorter lines before your calculation of distance functions. I saw the illustration in the paper that the start/end points of shorter line project on the longer one, but I believe a longer line can also project on the shorter one.
Things is, it is the line segs that are supposed to project on the encoding line
Did you priorly assume the shorter lines are the line segs? Cause that can lead to wrong results.

[apolcyn]

Thanks. I probably won't get a chance to take a good look at this soon (it's been a while). But with a quick look I believe that indeed this does make the assumption that the longer line is the encoding line, perhaps incorrectly. My quick guess is that in most cases, the partitioning phase should usually smooth things out so that the representative trajectory is longer than all line segments in the trajectory, and things can then work out.

I'd want to take make sure that the partitioning phase doesn't make that guarantee (that the representative trajectory isn't necessarily longer than all line segments) before making a change though (and find a test case where it fails), but I probably won't get around to this soon. But thanks for bringing this up.

[xiaoFine]

@apolcyn
I've been thinking over this issue and come out with a possible interpretation. The 'longer-shorter ' is necessary because it makes the distance functions symmetrical, as the symmetry is one of the necessary conditions of distance metric.
What do you think about it? :P