An Iterative Methodology for Unitary Quantum Channel Search

Bing-Ze Lu; Hao-Wei Huang; Matthew M. Lin

REVIEW 5 major objections 6 minor 20 references

An Iterative Methodology for Unitary Quantum Channel Search

T0 review · 5 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A polar-decomposition iteration recovers an unknown unitary quantum channel from limited input-output pairs.

desk verdict The polar-decomposition iteration is plausible, but the Section 5 recovery protocol uses non-physical quantum states and a sign error, so the paper's central practical claim fails as written. read the letter →

arxiv 2506.21455 v1 pith:ZU6GKD3H submitted 2025-06-26 math.NA cs.NAquant-ph

Matthew M. Lin , Hao-Wei Huang , Bing-Ze Lu This is my paper · ORCID

classification math.NAcs.NAquant-ph MSC 65F2515A2381P45

keywords unitaryquantumchannelpolardecompositionStiefelmanifoldFrobeniusnormstatetomographyreconstructionlocalminimumconvergenceanalysis

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims that an unknown unitary quantum channel—the reversible evolution that sends a state $\rho$ to $\sigma = U\rho U^*$—can be recovered from very few input-output pairs by solving a least-squares problem on the manifold of unitary matrices. The proposed iteration alternates a matrix multiplication with polar decomposition, $U^{(s+1)} = \mathrm{polar}(2\sigma U^{(s)}\rho)$, and the authors prove that the squared Frobenius misfit decreases and the iterates converge to a critical point, hence a local minimum, for almost all Hermitian positive definite pairs $(\sigma,\rho)$. They also prove that with one non-degenerate input state the candidate unitaries form an equivalence class differing by a diagonal unitary in the eigenbasis of $\rho$, and that two unitaries implementing the same channel differ only by a global phase. The reconstruction procedure uses $n^2+3n$ observables, which is less than the baseline $4n^2-2n-4$ Hermitian matrices needed for general unitary process tomography.

What carries the argument

The key object is the polar decomposition of the gradient-scaled product $2\sigma U^{(s)}\rho$, whose unitary factor is the nearest unitary matrix to that product in Frobenius norm. This makes the polar factor a natural projected-gradient step on the unitary group: the update is the unitary closest to the negative Euclidean gradient, so the fixed-point equation $U P = -\nabla g(U)$ exactly characterizes critical points of the constrained problem. The analysis is carried on the Stiefel manifold using its tangent-space geometry, expressed through skew-Hermitian coordinates, to compute projections, and the search-space reduction uses an equivalence relation $\#_V$ in which two solutions differ by a diagonal unitary in the eigenbasis $V$ of the non-degenerate input state $\rho$.

What would settle it

Run Algorithm 1 from $U^{(0)}=I$ with $\sigma=\operatorname{diag}(2,1)$ and $\rho=\operatorname{diag}(1,2)$: at the first step the matrix $I-(U^{(0)})^*U^{(1)}$ is generically not Hermitian positive semidefinite, so checking whether $g(U^{(1)})>g(U^{(0)})$ directly tests the claimed monotone-decrease guarantee.

Watch

Extended reading notes

Core claim

The central discovery is an iterative, first-order matrix algorithm that turns quantum channel identification into an optimization problem on the Stiefel manifold, the set of $n\times n$ unitary matrices. Starting from any unitary $U^{(0)}$, the update $U^{(s+1)} = \mathrm{polar}(2\sigma U^{(s)}\rho)$ is designed so that every fixed point satisfies the first-order optimality condition $U P = -\nabla g(U)$, and the paper argues that the objective $g(U)=\tfrac12\|\sigma - U\rho U^*\|_F^2$ never increases along the sequence and converges to a critical point for almost all data pairs. The reconstruction side shows that a single well-chosen non-degenerate state reduces the search to a diagonal phase matrix in the eigenbasis of the input state; probing the channel on $2n$ additional special states fixes those phases, yielding the full unitary up to a global phase with $n^2+3n$ tomography steps.

Load-bearing premise

The load-bearing premise is that $I-(U^{(s)})^*U^{(s+1)}$ is positive semidefinite at every iteration, but for two unitary matrices this matrix is generically not even Hermitian, so the monotone decrease and convergence to a critical point are not actually established.

Editorial extensions

If this is right

A single non-degenerate input-output pair already determines the channel up to a diagonal unitary in the eigenbasis of the input, so the effective search dimension drops from $n^2$ to $n$.
Two unitaries describing the same channel on all states differ only by a complex scalar of modulus 1; the global phase is the only remaining ambiguity.
Reconstructing the full unitary requires $n^2+3n$ observables, beating the $4n^2-2n-4$ Hermitian matrices required by generic unitary process tomography.
The iteration is implementable with standard matrix operations and converges numerically on random $10\times10$ channels and on an 8-dimensional composite gate made of Hadamard and CNOT gates.
Because the update uses only matrix products and a polar decomposition, the method is a first-order, dataset-light alternative to machine-learning-based channel identification.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the monotonicity argument is completed, the same update rule extends directly to multiple input-output pairs by minimizing a sum of squared Frobenius losses, since the polar-decomposition step is unchanged.
The equivalence-class reduction implies that after one non-degenerate pair only $n$ phase parameters remain to be measured, so a parallel hardware implementation of the $2n$ phase-probing steps could cut wall-clock time while keeping the $n^2+3n$ sample count.
The polar-factor step is a natural projected-gradient template that could be adapted to other matrix manifolds where the nearest feasible point is available in closed form, such as orthogonal or symplectic channel estimation.
The global-phase ambiguity proved for equivalent channel descriptions means downstream phase-sensitive algorithms must treat the overall phase as a gauge, while phase-insensitive uses can ignore it.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Punchline: the polar-decomposition iteration is a sensible idea and the convergence theorem is plausibly repairable, but the reconstruction protocol in Section 5 is not physically valid, so the paper's main practical claim does not hold as written.

What is new: the update by polar decomposition of 2σUρ is a natural first-order method on the unitary group, and I do not know a prior reference for exactly this update applied to unitary channel identification. The convergence analysis via Lemma 4.1 (closest unitary to a given matrix) is a reasonable strategy, and Theorem 2 (global phase uniqueness) is clean. The one-shot equivalence class theorem is standard but carefully stated. The numerics show the objective decreasing to machine precision, so there is a working algorithm underneath.

Soft spots: the stress test on Section 5 is right. The matrices (ρ_p,q,r)_± defined in Step 3 are Hermitian and trace one but not positive semidefinite; restricted to span{v_p,v_q}, (ρ_p,q,r)_+ is [[0,1/2],[1/2,0]] with eigenvalues ±1/2, so it is not a valid quantum input. The same holds for the minus sign. Passing a non-state through a channel is not a well-defined operation. Also, the text claims computing u_1* Φ(...) u_1 with U0 = D and u_1 = e_1 recovers d_pp d_qq*, but the expression is identically 1. So Steps 3–4 do not support the claimed n^2+3n-observable recovery.

The convergence proof also has a genuine flaw: Theorem 5 asserts I − (U^(s))*U^(s+1) is positive semidefinite, which is false for general unitaries. The descent inequality might still follow from Lemma 4.1 directly, so this is likely repairable. Lemma 4.2's diagonalization step contains algebra mistakes and assumes eigenvalues of ρ and σ are less than 1 without that being assumed anywhere. These are presentation-level defects compared with the Section 5 issue, but they are real.

Who this is for: someone working on manifold optimization for quantum process identification, or a numerical linear algebra audience interested in polar decomposition iterations. It is not a quantum information breakthrough; standard process tomography solves the identification problem with more data, and the paper does not compare against it.

Recommendation: send it out. The core algorithm and convergence analysis deserve a serious referee, and the Section 5 problem is concrete enough that a revision could fix it. If the reconstruction part cannot be repaired, the paper shrinks to a convergence analysis of a known-type iteration, which may still be publishable if honest about scope.

Referee Report

5 major / 6 minor

Summary. The paper proposes an iterative algorithm based on polar decomposition to identify a unitary quantum channel Φ(ρ)=UρU* from input-output pairs. It proves that the solutions of the one-pair problem form an equivalence class modulo a diagonal unitary (Theorem 1) and that channel representations are unique up to a global phase (Theorem 2). Algorithm 1 uses a polar-decomposition update; Theorems 5-6 claim monotone decrease of the objective and convergence to a critical point. Section 5 gives a reconstruction procedure to recover U using n^2+3n tomography steps. Numerical experiments on random 10×10 channels and a synthetic 8×8 quantum circuit are reported.

Significance. If the results were correct, the paper would offer a simple first-order polar-decomposition algorithm for unitary channel estimation with a convergence guarantee, plus a measurement-efficient reconstruction whose n^2+3n count improves on standard process tomography. The equivalence-class result (Theorem 1) and global-phase uniqueness (Theorem 2) are clean and appear correct. The numerical experiments, while not reproducible from the text, are consistent with convergence and accurate recovery. However, the advertised reconstruction protocol is not physically valid as written, and the convergence proofs contain incorrect algebraic assertions. The central idea is promising, but the manuscript in its current form does not rigorously support its headline claims.

major comments (5)

[Section 4, Theorem 5] The assertion in the proof of Theorem 5 that 'the matrix I − (U(s))* U(s+1) is positive semidefinite' is false for unitary matrices: for any eigenvalue e^{iθ} of U(s)*U(s+1), the corresponding eigenvalue of I − U(s)*U(s+1) is 1 − e^{iθ}, which is not real nonnegative unless θ=0. The proof of monotone decrease therefore does not hold as written. The claim itself appears repairable: from Lemma 4.1 one obtains Re Tr(σU(s)ρU(s+1)*) ≥ Re Tr(σU(s)ρU(s)*), and the remaining cross term Re Tr(σΔUρΔU*) is a nonnegative squared Frobenius norm. Please replace the false positive-semidefinite argument with this direct derivation.
[Section 4, Lemma 4.2] The diagonalization step contains algebraic errors. The displayed equality Re(Tr(QρΛρQρ* ΔU QσΛσQσ* ΔU*)) = Re(Tr((ΛσQσ*ΔU*Qρ)(ΛρQρ*ΔU*Qσ))) is incorrect: with the stated R = Qσ*ΔU*Qρ, the correct second factor should be ΛρR* = ΛρQρ*ΔUQσ, not ΛρQρ*ΔU*Qσ. The proof also invokes the identity 'tr(AB*) = tr(A*B)', which is false in general; only the real parts coincide for the Hermitian matrices considered. The lemma's conclusion may still be true, but the proof as written is not sound.
[Section 5, Step 3] The auxiliary matrices (ρ_{p,q,r})_± introduced in Step 3 are not density matrices. For example, in the subspace spanned by v_p and v_q, the matrix 1/2(v_p v_q* + v_q v_p*) has eigenvalues ±1/2, so the full matrix (which also contains v_r v_r*) has a negative eigenvalue for n≥3. Since Problem 1 and Algorithm 1 restrict to Hermitian positive definite inputs, these matrices cannot be prepared as quantum states or passed through the channel in the proposed protocol. The claimed reconstruction using n^2+3n quantum observables is therefore not a valid quantum-state procedure as stated.
[Section 5, Steps 2-4] The expression for the channel has the wrong placement of D: from U0 = U V D V* one gets U = U0 V D* V*, so Φ(ρ) = U0(V D* V*)ρ(V D V*)U0*, whereas the text writes U0(V D V*)ρ(V D* V*)U0*. Moreover, with the displayed formula and u1 equal to the first column of U0, the computation u1*Φ((ρ_{p,q,1})_±)u1 evaluates to 1 independently of D (take V=I and U0=D), so the claim that this yields d_pp d_qq* is not supported. The authors need to specify the correct measurement, e.g., cross terms u_p*Φ(ρ)u_q, and the correct ordering of D and D*.
[Abstract and Theorem 6] The abstract states that the algorithm identifies 'a critical point, which is also a local minimum,' but Theorem 6 only establishes convergence to a critical point; no second-order analysis is provided. In addition, the proof that the number of critical points is finite uses Lemma 4.5, but the first-order condition together with the unitary constraint is not a square polynomial system of the type covered by that lemma, and Lemma 4.5 only counts nonsingular solutions. Please either prove the local-minimum claim or revise the abstract, and provide a more careful finiteness argument.

minor comments (6)

[Section 2 and Lemma 3.1] The inner product ⟨X,Y⟩_R is defined with conjugation in equation (7) but as Σ x_ij y_ij (no conjugation) in the proof of Lemma 3.1; this inconsistency makes the derivative computation hard to verify.
[Section 2, equations (5)-(6)] The notation switches between Cm×m and Cn×n in the tangent-space formulas; please unify the dimensions.
[Section 4, Lemma 4.2] The comment that eigenvalues of ρ and σ are 'positive and strictly less than one' is not an assumption of Problem 1, which only requires positive definiteness; the inequality used is valid for all positive eigenvalues, so the condition is unnecessary.
[Section 5 and Section 6] There are several typos: 'p, q≠ i' should read 'p,q≠ r'; 'consided' should be 'considered'; in Example 2, 'result result' is duplicated.
[Section 6, Figures 1-2] The text states that the algorithm terminates after 1000 iterations, but the x-axes of the figures extend to 10^5; please clarify. Also, no random seed or code is provided, so the numerical experiments are not reproducible.
[Introduction] The novelty statement ('the use of matrix optimization methods to solve the quantum channel identification problem has not yet been explored') is too strong; there is existing work on optimizing over unitary matrices for quantum process learning. Please temper this claim and add references.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the derivation chain is self-contained, and the noted weaknesses are mathematical defects, not circular reasoning.

full rationale

The paper's claimed derivations are carried out from stated assumptions rather than from the conclusions being proved. Theorem 1 and Theorem 2 are direct linear-algebra proofs (diagonalization and commutation arguments), not restatements of the target results. The central algorithm is a polar-decomposition iteration whose monotonicity and convergence arguments rest on the externally cited polar-decomposition proximity lemma [2], on polynomial-system solution counts [1, 19], and on a real-analysis sequence lemma [5]; none of these are the authors' own prior results. The numerical experiments illustrate convergence but are not used to fit constants or to define the algorithm's output. The only self-citation, [4], supports the standard fact that the unitary group is an embedded submanifold of Euclidean space, and that fact is not load-bearing for the main claim. The genuine problems in the paper are correctness gaps rather than circularity: Theorem 5 asserts that I - (U^(s))*U^(s+1) is positive semidefinite, which is generally false for unitary matrices, and the Section 5 reconstruction uses matrices that are Hermitian and trace-one but not positive semidefinite, so they are not valid quantum states, with the written u1^*Phi((rho_p,q,1)_+)u1 computation evaluating to 1 under the stated conventions. These defects undermine the proof and the physical protocol, but they do not make the derivation equivalent to its inputs, so the circularity score is 0.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical entities or fitted constants. It relies on standard mathematical facts about polar decomposition and manifolds, plus domain assumptions about the noiselessness and non-degeneracy of the quantum states.

assumptions (4)

standard math The polar decomposition of a nonsingular matrix is unique and the unitary factor is the closest unitary matrix to it (Higham's lemma).
Used in Lemma 4.1 and throughout the algorithm.
standard math The tangent space of the unitary group at U is U times the skew-Hermitian matrices.
Used in Section 3 for the critical point characterization and in the derivation of the projection.
domain assumption The input and output states are exact and noiseless.
The algorithm and reconstruction assume sigma = U rho U* exactly; noise is not modeled.
domain assumption The selected state rho0 has non-degenerate eigenvalues.
Required for Theorem 1 to reduce the solution set to an equivalence class.

how reviews work

0 comments

Cite this review

Pith. "Pith review of An Iterative Methodology for Unitary Quantum Channel Search." pith.science (2026). https://pith.science/paper/ZU6GKD3H

@misc{pith2026250621455,
  author       = {Pith},
  title        = {Pith review of: An Iterative Methodology for Unitary Quantum Channel Search},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZU6GKD3H}},
  note         = {Machine review of arXiv:2506.21455}
}

read the original abstract

In this paper, we propose an iterative algorithm using polar decomposition to approximate a channel characterized by a single unitary matrix based on input-output quantum state pairs. In limited data, we state and prove that the optimal solution obtained from our method using one pair with a specific structure will generate an equivalent class, significantly reducing the dimension of the searching space. Furthermore, we prove that the unitary matrices describing the same channel differ by a complex number with modulus 1. We rigorously prove our proposed algorithm can ultimately identify a critical point, which is also a local minimum of the established objective function.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages

[1]

Sommese and Charles W

Andrew J. Sommese and Charles W. Wampler II, The numerical solution of systems of polynomials arising in engineering and science , World Scientific Publishing Co. Pte. Ltd., Hackensack, NJ, 2005, xxii+401 pages

work page 2005
[2]

Higham, Computing the polar decomposition—with applications , SIAM Journal on Scientific and Statistical Computing, 7(4):1160–1174, 1986

Nicholas J. Higham, Computing the polar decomposition—with applications , SIAM Journal on Scientific and Statistical Computing, 7(4):1160–1174, 1986

work page 1986
[3]

Arias, and Steven T

Alan Edelman, Tom´ as A. Arias, and Steven T. Smith, The geometry of algo- rithms with orthogonality constraints , SIAM Journal on Matrix Analysis and Applications, 20(2):303–353, 1998

work page 1998
[4]

Lin, Quantify- ing measurement-induced disturbance to distinguish correlations as classical or quantum, Journal of Scientific Computing, 99(1):Paper No

Yu-Chen Shu, Bing-Ze Lu, Kui-Yo Chen, and Matthew M. Lin, Quantify- ing measurement-induced disturbance to distinguish correlations as classical or quantum, Journal of Scientific Computing, 99(1):Paper No. 4, 15 pages, 2024

work page 2024
[5]

Chu and J

Moody T. Chu and J. Loren Watterson, On a multivariate eigenvalue prob- lem, Part I: Algebraic theory and a power method , SIAM Journal on Scientific Computing, 14(5):1089–1106, 1993

work page 1993
[6]

Huelga, Time evolution in closed quantum systems , Open Quantum Systems: An Introduction, Springer, 15–17, 2012

Angel Rivas and Susana F. Huelga, Time evolution in closed quantum systems , Open Quantum Systems: An Introduction, Springer, 15–17, 2012

work page 2012
[7]

Raban Iten, Roger Colbeck, and Matthias Christandl, Quantum circuits for quantum channels , Physical Review A, 95(5):052316, 2017

work page 2017
[8]

Flammia, Stephen Becker, and Jens Eis- ert, Quantum state tomography via compressed sensing , Physical Review Letters, 105(15):150401, 2010

David Gross, Yi-Kai Liu, Steven T. Flammia, Stephen Becker, and Jens Eis- ert, Quantum state tomography via compressed sensing , Physical Review Letters, 105(15):150401, 2010

work page 2010

Show all 20 references

[9]

Chuang and Michael A

Isaac L. Chuang and Michael A. Nielsen, Prescription for experimental deter- mination of the dynamics of a quantum black box , Journal of Modern Optics, 44(11-12):2455–2467, 1997

1997
[10]

Tran, Daniel Carney, and Jacob M

Jonathan Kunjummen, Minh C. Tran, Daniel Carney, and Jacob M. Tay- lor, Shadow process tomography of quantum channels , Physical Review A, 107(4):042403, 2023. 19

2023
[11]

Nicolas Gisin and Rob Thew, Quantum communication , Nature Photonics, 1(3):165–171, 2007

2007
[12]

Petersen, Quantum control theory and applications: a survey, IET Control Theory & Applications, 4(12):2651–2671, 2010

Daoyi Dong and Ian R. Petersen, Quantum control theory and applications: a survey, IET Control Theory & Applications, 4(12):2651–2671, 2010

2010
[13]

D´ enes Petz, Sufficient subalgebras and the relative entropy of states of a von Neumann algebra, Communications in Mathematical Physics, 105:123–131, 1986

1986
[14]

Alessandro Bisio, Giulio Chiribella, Giacomo Mauro D’Ariano, Stefano Facchini, and Paolo Perinotti, Optimal quantum learning of a unitary transformation , Physical Review A, 81(3):032324, 2010

2010
[15]

Albert, and John Preskill1, Provably efficient machine learning for quantum many-body problems

Hsin-Yuan Huang, Richard Kueng, Giacomo Torlai, Victor V. Albert, and John Preskill1, Provably efficient machine learning for quantum many-body problems. Science 377.6613 (2022): eabk3333

2022
[16]

Mart ´ ın-Guerrero, and Enrique Solano, Supervised quantum learning without measurements, Scientific Reports, 7(1):13645, 2017

Unai Alvarez-Rodriguez, Lucas Lamata, Pablo Escandell-Montero, Jos´ e D. Mart ´ ın-Guerrero, and Enrique Solano, Supervised quantum learning without measurements, Scientific Reports, 7(1):13645, 2017

2017
[17]

Giovanni Cemin, Marcel Cech, Erik Weiss, Stanislaw Soltan, Daniel Braun, Igor Lesanovsky, and Federico Carollo, Machine learning of quantum channels on NISQ devices, arXiv preprint arXiv:2405.12598, 2024

2024 arXiv
[18]

A. S. Kardashin, A. V. Vlasova, A. A. Pervishko, Dmitry Yudin, and Jacob D. Biamonte, Quantum-machine-learning channel discrimination , Physical Review A, 106(3):032409, 2022

2022
[19]

Yu Guan, Nan Jiang, Bo Dong, and Moody T. Chu, Convergence analy- sis of alternating direction methods: A general framework and its applications to tensor approximations , Available at https://mtchu.math.ncsu.edu/Research/ Papers/convergence general-MP.pdf, 2018

2018
[20]

https: //doi.org/10.1063/1.4867625 20

Gus Gutoski and Nathaniel Johnston, Process tomography for unitary quantum channels, Journal of Mathematical Physics, 1 March 2014; 55 (3): 032201. https: //doi.org/10.1063/1.4867625 20

2014 doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Sommese and Charles W

Andrew J. Sommese and Charles W. Wampler II, The numerical solution of systems of polynomials arising in engineering and science , World Scientific Publishing Co. Pte. Ltd., Hackensack, NJ, 2005, xxii+401 pages

work page 2005

[2] [2]

Higham, Computing the polar decomposition—with applications , SIAM Journal on Scientific and Statistical Computing, 7(4):1160–1174, 1986

Nicholas J. Higham, Computing the polar decomposition—with applications , SIAM Journal on Scientific and Statistical Computing, 7(4):1160–1174, 1986

work page 1986

[3] [3]

Arias, and Steven T

Alan Edelman, Tom´ as A. Arias, and Steven T. Smith, The geometry of algo- rithms with orthogonality constraints , SIAM Journal on Matrix Analysis and Applications, 20(2):303–353, 1998

work page 1998

[4] [4]

Lin, Quantify- ing measurement-induced disturbance to distinguish correlations as classical or quantum, Journal of Scientific Computing, 99(1):Paper No

Yu-Chen Shu, Bing-Ze Lu, Kui-Yo Chen, and Matthew M. Lin, Quantify- ing measurement-induced disturbance to distinguish correlations as classical or quantum, Journal of Scientific Computing, 99(1):Paper No. 4, 15 pages, 2024

work page 2024

[5] [5]

Chu and J

Moody T. Chu and J. Loren Watterson, On a multivariate eigenvalue prob- lem, Part I: Algebraic theory and a power method , SIAM Journal on Scientific Computing, 14(5):1089–1106, 1993

work page 1993

[6] [6]

Huelga, Time evolution in closed quantum systems , Open Quantum Systems: An Introduction, Springer, 15–17, 2012

Angel Rivas and Susana F. Huelga, Time evolution in closed quantum systems , Open Quantum Systems: An Introduction, Springer, 15–17, 2012

work page 2012

[7] [7]

Raban Iten, Roger Colbeck, and Matthias Christandl, Quantum circuits for quantum channels , Physical Review A, 95(5):052316, 2017

work page 2017

[8] [8]

Flammia, Stephen Becker, and Jens Eis- ert, Quantum state tomography via compressed sensing , Physical Review Letters, 105(15):150401, 2010

David Gross, Yi-Kai Liu, Steven T. Flammia, Stephen Becker, and Jens Eis- ert, Quantum state tomography via compressed sensing , Physical Review Letters, 105(15):150401, 2010

work page 2010