I have a question somewhat in line with the one asked here. That is, I am interested in how the persistent homology for a sublevel set of a function ($\{x \: : \: f(x) \leq c\}$) is computed. For finite point clouds it is my understanding that a filtration of complexes is constructed directly from the point cloud, which is then used to construct the persistent homology.
But a sublevel set may not be finite, so how do you compute it's persistent homology? In the aforementioned post, someone describes a strategy where a grid is placed over the entire input space, and grid points lying inside the sublevel set are used to construct a filtration of complexes, from which the persistent homology can be computed (if I am understanding correctly).
First of all, this method seems like it would depend exponentially on the dimension of the input space.
Are there faster approaches for high dimensional spaces?
Second of all, how is this approach theoretically justified? I can see intuitively why such an approximation would work once the grid is fine enough, but wondered if there is a rigorous explanation. For example, is there a theorem that says once the grid is fine enough that some complex in the filtration will be homeomorphic to the sublevel set?
Best Answer