Finite Shot Quantum Moment Estimation

---
name: finite-shot-quantum-moment-estimation
description: "Finite-shot moment estimation methodology for quantum metrology — bias-corrected estimators, calibration curve analysis, and sensitivity corrections for quantum parameter estimation beyond the Cramer-Rao bound. Use when working with quantum metrology, parameter estimation, moment estimators, quantum Cramer-Rao bound, finite measurement number, bias correction, or calibration curves in quantum sensing and quantum information tasks."
metadata:
  arxiv_id: "2606.25920"
  published: "2026-06-24"
  authors: "Shaowei Du, Shuheng Liu, Weidong Li, Luca Pezze, Augusto Smerzi, Qiongyi He"
  tags: [quantum-metrology, parameter-estimation, moment-estimation, bias-correction, cramer-rao, quantum-sensing]
---

# Finite-Shot Quantum Moment Estimation

Methodology from arXiv:2606.25920 for finite-measurement theory of method-of-moments estimation in quantum metrology.

## Core Problem

The quantum Cramer-Rao bound (QCRB) is asymptotic only — it doesn't specify how many measurements (ν) are needed for a concrete estimator to approach it.

## Key Results

1. **Bias-corrected estimator**: Nonlinear calibration curves make moment estimators biased at finite ν. Construct bias-corrected estimator with bias O(ν⁻³).

2. **Density-matrix condition**: General condition under which the full 1/ν² correction vanishes.

3. **Unitary examples**: Leading residual correction at O(1/ν³), governed by calibration curvature, reducible by higher-rank components.

4. **Measurement thresholds**: Quantify how many measurements needed before asymptotic sensitivity becomes operationally visible.

## Mathematical Framework

### Calibration Curve

For a calibrating observable M and parameter θ:

```
m(θ) = Tr[M ρ(θ)]
```

The moment estimator solves: `m̂ = m(θ̂)` where m̂ is sample mean.

### Bias Correction

Standard moment estimator bias at finite ν:

```
Bias(θ̂) = (1/2ν) * m''(θ) / [m'(θ)]² * Var(M) + O(ν⁻²)
```

Bias-corrected estimator eliminates O(ν⁻¹) and O(ν⁻²) terms, achieving O(ν⁻³).

### Sensitivity Expansion

```
Δθ² = (ΔM)² / [ν · m'(θ)²] + c₂/ν² + c₃/ν³ + ...
```

Where coefficients cₙ depend on calibration curve derivatives and central moments of M.

## Usage Patterns

### When to Apply

- Quantum parameter estimation with limited measurement budget
- Moment-based estimation where calibration is nonlinear
- Quantum sensing protocols needing finite-shot sensitivity guarantees
- Beyond-QCRB analysis for concrete estimator performance

### When NOT to Apply

- Full likelihood estimation (MLE) already implemented
- Linear calibration curves (standard error propagation suffices)
- Asymptotic regime where ν → ∞

## Practical Implementation

```python
def bias_corrected_moment_estimator(sample_mean, calib_func, calib_deriv1, calib_deriv2, n_shots):
    """
    sample_mean: observed sample mean of observable M
    calib_func: m(θ) = Tr[M ρ(θ)]
    calib_deriv1: m'(θ)
    calib_deriv2: m''(θ)
    n_shots: number of measurements ν
    """
    # Inverse calibration curve
    theta_naive = inverse_calib(sample_mean)
    
    # Bias correction term
    var_m = compute_variance(sample_mean, n_shots)
    bias = (1 / (2 * n_shots)) * calib_deriv2(theta_naive) * var_m / calib_deriv1(theta_naive)**2
    
    return theta_naive - bias
```

## Pitfalls

- **Calibration curve inversion**: Requires monotonic m(θ); non-monotonic curves need piecewise treatment
- **Higher-rank observables**: May need to decompose M into components to reduce curvature effects
- **Density-matrix singularity**: Bias vanishing condition requires checking density-matrix rank