Subsampling Confidence Bound for Persistent Diagram via Time-delay Embedding (2512.06324v1)

Abstract: Time-delay embedding is a fundamental technique in Topological Data Analysis (TDA) for reconstructing the phase space dynamics of time-series data. While persistent homology effectively identifies topological features, such as cycles associated with periodicity, a rigorous statistical framework for quantifying the uncertainty of these features has been lacking in this context. In this paper, we propose a subsampling based method to construct confidence sets for persistence diagrams derived from time-delay embeddings. We establish finite sample guarantees for the validity of these confidence bounds under regularity conditions specifically for $C^{1,1}$ functions with positive reach and prove their asymptotic convergence as the embedding dimension tends to infinity. This framework provides a principled statistical test for periodicity, enabling the distinction between true periodic signals and non-periodic approximations. Simulation studies demonstrate that our method achieves detection performance comparable to the Generalized Lomb-Scargle periodogram on periodic data while exhibiting superior robustness in distinguishing non-periodic signals with time-varying frequencies, such as chirp signals.