machine learning - Suggestions for multiple DTW cost paths?

Question

While writing a Dynamic Time Warping algorithm from scratch in MaxMSP, I stumbled upon an issue I didn't anticipate. During the minimum cost path search, there might be two, or even three, surrounding cell neighbors with the same minimum value. In some cases, when the neighbors are adjacent, it's likely that the choice between either one is close to insignificant. However, when that's not the case, the resulting cost path may differ greatly.

For the sake of clarity, here's purposefully inconvenient scenario: multiple-path cost matrix

My questions on this issue are the following:

1. Does this problem have a common name within the literature? I tried looking it up but couldn't find much information.
2. What solutions or alternatives have been implemented so far to account for this scenario? The first thing that comes to mind would be like incorporating a 'look ahead' function, where you look for the minimum neighbors of every repeated minimum value, and which ever has the smallest neighbor, that one becomes part of the final cost path.

question from:https://stackoverflow.com/questions/66057014/suggestions-for-multiple-dtw-cost-paths

Answer 1

Waitting for answers