5. Combinatorial aspect

We've presented the context of multi-track merging, a relatively lengthy exercise given the function's downstream position, but one that is necessary for overall understanding. We'll now turn to the algorithmic aspects. The problem to be addressed is known in statistics as unsupervised classification: the aim is to associate tracks into classes (i.e. into groups) without knowing either the number of these classes, or the tracks making them up. The approach adopted here is known as "hard-decision", since a track can belong to only one group. As in tracking (§ 3