International Conference on Federated Learning and Intelligent Computing Systems (FLICS)
Quantum Data Statistics: An Operator-Theoretic Framework for Structured Uncertainty
Pedro Caceres, Zeraoulia Rafik
Abstract
We introduce Quantum Data Statistics (QDS), an operator-theoretic framework for representing structured uncertainty in time-evolving empirical systems. QDS maps each empirical distribution to a density operator on a finite-dimensional Hilbert space, enabling diagnostics that are sensitive to how uncertainty is organized, not only how much uncertainty is present. The operator state admits a spectral decomposition whose eigenvalues and eigenvectors reveal latent “regime” structure; off-diagonal entries quantify interaction (coherence) between observational components; and a dual-domain uncertainty functional captures simultaneous localization in observation and spectral coordinates. We also define geometric descriptors (curvature and instability) of the statistical trajectory. In a quantum-inspired logistic-map benchmark spanning fixed-point, periodic, and chaotic regimes, QDS invariants sharply localize bifurcations and provide earlywarning signals that outperform classical variance and Lyapunovbased indicators in short windows.