Compressive Quantum Tomography

---
name: compressive-quantum-tomography
description: "Compressive quantum state tomography methodology — unified framework for structured quantum state recovery using low-rankness, tensor networks, and compressive sensing principles. Bridges statistics, optimization, and quantum information theory for scalable quantum state characterization."
---

# Compressive Quantum Tomography

## Description
Quantum state tomography (QST) is a fundamental task in quantum information science that aims to reconstruct unknown quantum states from measurement data. The exponential growth of Hilbert-space dimension with system size makes full tomography statistically and computationally prohibitive. This skill covers compressive and structured approaches that exploit prior structure — low-rankness, tensor-network representations, shallow quantum circuits, and neural quantum states — to substantially reduce effective degrees of freedom and enable scalable recovery.

## Activation Keywords
- compressive quantum tomography
- structured quantum state tomography
- quantum state recovery
- low-rank quantum tomography
- matrix sensing quantum
- compressive sensing quantum
- quantum measurement design
- 压缩量子层析
- 结构化量子态重建
- quantum tomography sample complexity
- randomized measurements quantum
- POVM quantum tomography
- neural quantum states tomography
- tensor network quantum state
- informationally complete POVM

## Tools Used
- terminal: Run quantum simulation and optimization scripts
- write_file: Create QST reconstruction code
- read_file: Read measurement data and configuration files
- search_files: Find related quantum tomography implementations

## Core Concepts

### 1. Compact State Representations
Structured quantum states have far fewer degrees of freedom than the full 2^n - 1 parameters:
- **Low-rank states**: Rank-r density matrices require O(r * 2^n) parameters instead of O(4^n)
- **Tensor network states**: MPS/PEPS representations with polynomial parameter scaling
- **Shallow circuit states**: States preparable by depth-d circuits with O(d * n) parameters
- **Neural quantum states**: Neural network parameterizations (RBM, autoregressive models)

### 2. Measurement Design
- **Informationally complete POVMs**: Minimum measurements to uniquely identify any quantum state
- **Randomized measurements**: Haar-random or locally random unitary rotations + computational basis measurement
- **Geometric preservation**: RIP (Restricted Isometry Property) and related conditions guaranteeing stable recovery
- **Sample complexity**: Bounds on number of measurements needed as function of state complexity

### 3. Computational Algorithms
- **Convex optimization**: Nuclear norm minimization, trace norm regularization
- **Non-convex optimization**: Gradient descent on low-rank factorizations
- **Compressive sensing**: L1-minimization, iterative hard thresholding
- **Matrix/tensor sensing**: Extensions of classical compressive sensing to quantum domains

## Usage Patterns

### Pattern 1: Low-Rank Quantum State Recovery
When the target state is approximately pure or low-rank:
1. Design randomized measurement scheme (random Clifford or local random unitaries)
2. Collect measurement statistics
3. Solve nuclear norm minimization: min ||ρ||_1 s.t. measurement constraints
4. Or use factorized approach: ρ = XX†, optimize over X directly

### Pattern 2: Tensor Network State Tomography
For states with limited entanglement (area law states):
1. Choose tensor network ansatz (MPS for 1D, PEPS for 2D)
2. Design local measurement scheme
3. Use alternating least squares or gradient-based optimization
4. Validate with cross-entropy or fidelity estimation

### Pattern 3: Neural Quantum State Tomography
For complex many-body states:
1. Choose neural architecture (RBM, CNN, autoregressive Transformer)
2. Generate training data from measurements
3. Train via maximum likelihood or variational methods
4. Extract physical observables from trained model

### Pattern 4: Sample Complexity Analysis
When designing experiments:
1. Identify state structure (rank, tensor rank, circuit depth)
2. Apply theoretical bounds for measurement complexity
3. Design measurement scheme matching the bound
4. Validate reconstruction quality with held-out measurements

## Instructions for Agents

### Step 1: Characterize State Structure
Determine what structure the target quantum state likely has:
- Is it approximately pure? → Low-rank methods
- Does it have limited entanglement? → Tensor network methods
- Is it generated by a shallow circuit? → Circuit-based tomography
- Is it a thermal state? → Gibbs state tomography

### Step 2: Select Measurement Scheme
Choose measurements that preserve the structure:
- **Low-rank**: Pauli measurements, random Clifford measurements
- **Tensor network**: Local measurements on subsystems
- **General**: Informationally complete POVMs with geometric preservation guarantees

### Step 3: Choose Recovery Algorithm
Match algorithm to structure and measurement scheme:
- **Convex**: Nuclear norm minimization (guaranteed but slow for large systems)
- **Non-convex**: Riemannian optimization on low-rank manifolds (fast, local minima risk)
- **Iterative**: Hard thresholding, alternating projections (simple, moderate guarantees)

### Step 4: Validate Reconstruction
- Compute fidelity with known states (benchmark)
- Check prediction accuracy on held-out measurements
- Verify physical constraints (positivity, trace = 1)
- Estimate error bounds using concentration inequalities

## Error Handling

### Insufficient Measurements
- **Symptom**: Reconstruction fails to converge or produces unphysical states
- **Fix**: Increase measurement count, verify measurement scheme is informationally complete
- **Rule of thumb**: O(r * 2^n * log(2^n)) measurements for rank-r states

### Unphysical States
- **Symptom**: Reconstructed density matrix has negative eigenvalues
- **Fix**: Enforce positivity constraints, use convex formulations, or apply nearest PSD projection

### Scalability Issues
- **Symptom**: Algorithms fail for n > 20 qubits
- **Fix**: Switch to tensor network or neural quantum state representations, use distributed optimization

## Examples

### Example 1: Low-Rank State Tomography with Compressive Sensing
```python
# Conceptual workflow for compressive QST
import numpy as np
from scipy.optimize import minimize

# Given: measurement outcomes M_i = Tr(O_i * ρ_true) for i = 1,...,m
# Goal: recover ρ_true assuming it's approximately rank-r

def nuclear_norm_minimization(measurements, operators, n_qubits):
    """Recover low-rank quantum state from compressive measurements."""
    dim = 2**n_qubits
    
    # Initialize with maximally mixed state
    rho_init = np.eye(dim) / dim
    
    # Nuclear norm minimization via convex optimization
    def objective(rho_flat):
        rho = rho_flat.reshape(dim, dim).view(np.complex128)
        return np.sum(np.abs(np.linalg.eigvalsh(rho)))  # nuclear norm
    
    def constraints(rho_flat):
        rho = rho_flat.reshape(dim, dim).view(np.complex128)
        return [np.abs(np.trace(op @ rho) - outcome) for op, outcome in zip(operators, measurements)]
    
    # Solve and project to PSD
    result = minimize(objective, rho_init.flatten(), constraints=constraints)
    rho_recovered = project_psd(result.x.reshape(dim, dim))
    return rho_recovered
```

### Example 2: Randomized Measurement Protocol
```python
def design_randomized_measurements(n_qubits, n_measurements):
    """Design randomized measurement scheme for compressive QST."""
    measurements = []
    for _ in range(n_measurements):
        # Random local unitary + computational basis measurement
        U = random_local_unitary(n_qubits)
        measurements.append(U)
    return measurements

def collect_data(state, measurement_unitaries):
    """Simulate collecting measurement data."""
    outcomes = []
    for U in measurement_unitaries:
        # Apply U, measure in computational basis
        rotated = U @ state @ U.conj().T
        prob_diag = np.diag(rotated).real
        outcome = np.random.choice(len(prob_diag), p=prob_diag)
        outcomes.append(outcome)
    return outcomes
```

## Resources
- arXiv: 2605.27191 — "Statistical and Algorithmic Foundations of Probing Quantum Systems with Compressive Measurements: A Review"
- Related: `quantum-state-preparation-medical`, `qml-feature-encoding`, `quantum-ml-data-loading`
- Key references: Candes & Tao (compressive sensing), Gross et al. (quantum tomography), Huang et al. (shadow tomography)

## Related Skills
- quantum-state-preparation-nn: Neural network quantum state preparation
- qml-feature-encoding: Quantum feature encoding methods
- quantum-fisher-information-duality: QFI bounds for parameter estimation
- tomography-by-design: Algebraic approach to low-rank quantum states