During the topological analysis, we encountered a saddle point on the surface, which complicated the classification because every neighbourhood of it contained points on both sides of its tangent plane.

In the optimization lecture, the professor explained how saddle points can slow down the convergence of gradient-based algorithms.