Core Concepts
The authors demonstrate the NP-hardness of computing p-presentation distances for merge trees and t-parameter persistence modules, providing a novel strategy applicable to various settings in topological data analysis.
Abstract
The content discusses the introduction of p-presentation distances for merge trees and multiparameter persistence modules, highlighting their sensitivity compared to interleaving distances. The authors prove the NP-hardness of computing these distances for various values of p, showcasing a novel approach that can be extended to other distance metrics based on sums or p-norms. Merge trees and persistence modules are fundamental concepts in topological data analysis, with applications in data visualization and homology studies. The article presents a detailed proof strategy involving algebraic and combinatorial objects, emphasizing the complexity of computing these distances. The study also acknowledges contributions from related research areas and provides an outline for further exploration.
Stats
It is well-known that computing the interleaving distance is NP-hard in both cases.
Computing the p-presentation distance is NP-hard for all p ∈ [1, ∞) for both merge trees and t-parameter persistence modules for any t ≥ 2.
For single-parameter persistence, the interleaving distance is equal to the bottleneck distance.
Recently, there has been an increasing interest in studying the homology of filtrations dependent on more than a single variable.
These distances are called p-presentation distances, and it was shown that these distances are all universal in their respective settings.
The problems of computing the p-presentation distance for merge trees and t-parameter persistence modules for any t ≥ 2 are NP-hard for all p ∈ [1, ∞).
Our two proofs are summarized in Theorem 1.1.