Two passengers boarded an empty car on a circle-line subway. After a full loop, they ended up sitting in each other's seats — even though neither moved. The same confusion happens in Dynamical Systems when Machine Learning tracks objects along a closed path. The mathematical method for this was developed by Henri Poincaré. A Graphs is built to record the births and deaths of features as the scale changes. If the path is closed, the labels can be permuted after a full loop.

To find the cause, researchers looked at information theory about the order of births and deaths. Geometry holds the key: every closed curve has a “skeleton”—the set of centers of circles that touch the curve at two points. This skeleton consists of smooth lines whose motion follows differential equations. Sometimes, two of these lines cross. Around such an intersection, space divides into four sectors, like a clock face. Spectral Analysis showed that only this type of crossing causes permutations. numerical simulation confirmed it: all cases of confusion are tied to this point.

The discovery offers a way to optimization algorithms: now we can design loops that either avoid the special point or intentionally capture it, giving us control over the permutations. In nonlinear dynamics this helps keep objects from getting mixed up. Back to the subway: the special point is like a transfer station where an invisible mechanism swaps passengers. Knowing it lets us plan routes with no surprises.