Degree-dependent minimax laws for Gaussian factorial moments
Loading original first page…
Abstract
We classify the copy budgets for estimating photon-number factorial moments of an unknown centered single-mode Gaussian state. Both occupation and complex anomalous covariance are unknown, and arbitrary bias is allowed. For every degree n ≥ 2, collective quantum procedures and standard heterodyne have minimax squared relative risk of order min{1, (1 + a⁻¹)²n²/N}, uniformly in degree, copy count and positive occupation floor a. Degree one is exceptional: its quantum scale is (1 + a⁻¹)/N. We prove complete endpoint classifications: risk tends to zero exactly when its resource scale vanishes, and for n ≥ 2 tends to one exactly when that scale diverges. Bounded positive scales keep both risks away from zero and one. Nevertheless one scalar resource scale cannot determine all critical risks. A growing-degree boundary experiment has a Poisson observation with a modified-Bessel target; its exact classical minimax risk yields an explicit quantum lower. At the same occupation floor and exactly the same resource scale, this lower separates from a published quartic upper. The universal copy-budget order thus coexists with genuinely degree-dependent critical behavior.
Read the full paper in the original PDF.
Edition files
- main.pdf124.1 KiB · pdf
SHA-256
6b2b7c211fcbad907a5fe52dde9a2b303b76bf895dc9b9d8b2eac37e6cd4160e - Gaussian_hierarchy_publication.zip137.9 KiB · package
SHA-256
9fdcf0c5a9aebea142fec2acc112ccedee171945f4f4ddd6b0f287a334949e79
Available editions
Research details
- Subjects
- Mathematics · Quantum Information Theory · Statistics and Data Science · Statistical Theory
- Keywords
- Gaussian quantum states, minimax estimation, relative risk
- Manuscript date
- 2026-11-01
- Reader reuse
- none
Scope
We classify the copy budgets for estimating photon-number factorial moments of an unknown centered single-mode Gaussian state.
Disclosures
OpenAI Codex was used for mathematical, computational and editorial assistance in preparing this work.
Reproduction
The source package contains the complete editable manuscript, build instructions, exact finite reproducibility controls, citation metadata and member checksums.