##  [Maximum Likelihood Estimation (Tomography)](/maximum-likelihood-estimation-tomography-0) 

 Definition

A measurement concept defining how outcomes are modeled and how state descriptions are updated after an outcome is recorded. It governs outcome probabilities, information extraction, and the disturbance introduced by the measurement interaction. It does not yield reliable inference without adequate calibration, sufficient data, and appropriate estimation procedures. It supports reconstruction and validation of state and process descriptions from experimental statistics. The concept is generally stable, though practical implementations and estimation methods evolve over time.



 

 

 

 

 

 





## Principle

Principle

Estimate the parameter (the density matrix) by maximizing the likelihood function L(ρ)=Pr(data|ρ) over the set of physically allowed states; enforce positivity and trace one so the solution is a valid density operator. Numerical convex or constrained optimization methods implement the principle when closed-form solutions are unavailable.

 

 

 

 

 





## Demonstration

Demonstration

Reconstructing a single-qubit state from counts of Pauli-basis measurements: given outcome counts for σx, σy, σz, form the likelihood of observed counts as a function of the Bloch vector components, maximize that likelihood under the constraint that the Bloch vector length ≤1 to obtain the MLE density matrix.

 

 

 

 

## Misapplication

Misapplication

Treating the raw maximizer without enforcing positivity (yielding an unphysical matrix), assuming the MLE is unbiased for finite sample sizes, or reporting MLE without quantifying uncertainty. Using MLE with an incorrect measurement model (wrong POVM) gives systematically wrong states.

 

 

 

 

 





## Consequence

Consequence

Produces a physically valid point estimate that is asymptotically efficient (achieves classical Fisher-information limits for large data) and often yields high-fidelity reconstructions; finite-sample estimates can be biased and require resampling, Fisher-information evaluation, or Bayesian intervals to quantify uncertainty.

 

 

 

 

## Reversal

Reversal

Bayesian tomography produces a posterior distribution over states instead of a single maximizer; the Maximum A Posteriori (MAP) estimate resembles MLE when the prior is flat, but Bayesian methods emphasize uncertainty and prior regularization.

 

 

 

 

 





## Boundary

Boundary

Applies when a well-specified measurement model and sufficient data exist and when one seeks a point estimate for the density operator. It excludes tasks requiring full posterior uncertainty without extra analysis, ill-posed incomplete measurement sets where likelihood is flat over a manifold, and contexts where model mismatch invalidates the likelihood.

 

 

 

 

 





## Semantic Tension

Semantic Tension

Tension between point-estimate pragmatism (MLE) and full uncertainty representation (Bayesian); between enforcing physical constraints (which can introduce bias) and preserving unbiasedness in the classical estimator sense; and between numerical tractability and global optimality in high-dimensional Hilbert spaces.

 

 

 

 

 





## Synthesis

Synthesis

Maximum-likelihood tomography is the constrained optimization approach that returns the physically valid density matrix most consistent with observed measurement frequencies; it is efficient and practical for many datasets but demands careful enforcement of physical constraints and additional methods for reliable uncertainty quantification.