.png){kind=icon}
.png){kind=icon}
Return
Functional Data Representation with Merge Trees
.png){kind=icon}
PRE
AI
DOI：10.1080/10618600.2026.2654769.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In this paper, we address the problem of representing functional data using tools from algebraic topology. We model functions through merge trees, which, like the more widely used persistence diagrams, are invariant under homeomorphic reparameterizations of their domain. This invariance allows statistical analyses that are robust to functional misalignment. We study a recently introduced metric for merge trees, establish its stability properties, and prove the consistency and convergence rates of a corresponding class of estimators. To illustrate the advantages of our topological approach to functional data analysis, we apply it to the AneuRisk65 dataset, replicating a well-known supervised classification task that made this dataset a benchmark for misaligned functional data. We also compare our results with those obtained using the Fisher-Rao metric, which further supports the effectiveness of the proposed method. The s includes an extensive comparison between merge trees and persistence diagrams, outlining their similarities and differences to help analysts choose the most suitable representation for a given application.
Keywords:
Consistency of merge trees
Functional alignment
Functional data analysis
Kernel estimators
Topological data analysis
Tree edit distance
Journal
J
IF:
1.8
Papers:
138
Citations:
6.4K