Recovering Hidden Structure from Noisy Observations via Phase-Space Reconstruction

This work explores the recovery of hidden dynamical structure from noisy observational data using phase-space embedding methods. An analogy is drawn between event-level neutrino tagging and nonlinear state-space reconstruction, highlighting how representation choices influence observable physical structure. The approach is demonstrated using pulsar data from the European Pulsar Network, comparing Fourier spectral methods with geometric reconstruction in phase space. The results suggest that structural information can remain preserved even under significant observational noise, supporting the view that apparent information loss may be representation-dependent. This preprint corresponds to a journal submission currently under review.