Pith. sign in

REVIEW 5 major objections 5 minor 1 cited by

A Unified Blockwise Measurement Design for Learning Quantum Channels and Lindbladians via Low-Rank Matrix Sensing

T0 review · 5 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read This paper claims that any low-rank quantum superoperator, including both quantum channels and Lindbladians, can be recovered from O(r N^2 log^6 N) random Pauli measurements, and that a blockwise first-row design achieves the same scaling…

desk verdict A genuinely useful blockwise measurement design for low-rank quantum superoperator learning, with printed theory that needs several honest fixes before its guarantees can be quoted as-is. read the letter →

arxiv 2501.14080 v1 pith:45JO2EO5 submitted 2025-01-23 quant-ph stat.ML

Quanjun Lang , Jianfeng Lu This is my paper · ORCID
classification quant-phstat.ML
keywords quantumprocesstomographyLindbladianlearninglow-rankmatrixsensingrestrictedisometrypropertyPaulimeasurementsalternatingleastsquareschannelblockwisemeasurementdesign
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 tries to establish that an unknown low-rank quantum superoperator, written as $K\rho = \sum_{k=1}^{r_+} V_k \rho V_k^\dagger - \sum_{k=1}^{r_-} U_k \rho U_k^\dagger$, can be reconstructed from far fewer measurements than full quantum process tomography. The authors show that roughly $O(r N^2 \log^6 N)$ random Pauli-type measurements suffice, with explicit recovery guarantees in the presence of noise, and that this scaling holds for quantum channels (where $r_- = 0$) and for Lindbladians (where the reshaped matrix carries negative eigenvalues). They then propose a blockwise design that learns only the first row of $N \times N$ blocks of the reshaped operator and reconstructs the rest deterministically. A sympathetic reader would care because this unifies two tasks, quantum channel tomography and Lindbladian learning, under one low-rank matrix sensing framework with a tractable optimization landscape.

What carries the argument

The load-bearing objects are the reshaping operator $R$, which converts the superoperator into a Hermitian matrix $K$ with possibly negative eigenvalues, and the first-row blockwise sensing map. $R$ maps Kronecker products to rank-one outer products, giving $K = VV^\dagger - UU^\dagger$, which is what lets matrix sensing apply to Lindbladians. Measurement is expressed as $\langle \rho_1, O \rangle = \langle \rho_0 \otimes O, K \rangle$, and the invalid input states $E_{lk}$ are emulated by four valid density matrices. Random Pauli observables make the sensing map satisfy the restricted isometry property, and the landscape theorem imported from the matrix sensing literature then guarantees that all local minima of the regularized objective recover the true matrix. The blockwise design uses the same observables for all initial states, so the stacked operator inherits the RIP, alternating least squares with Nesterov acceleration solves the factored subproblems, and the final deterministic step uses randomized SVD of the first row plus the pseudo-inverse of $J_1^\dagger$ to complete the full matrix.

What would settle it

Take a rank-2 reshaped superoperator $K^*$ that is diagonal in the canonical basis with two nonzero entries, generate noiseless Pauli measurements from it, run the blockwise pipeline (first-row ALS plus Algorithm 5), and check whether the relative Frobenius error fails to reach machine precision while the random-design ALS on the same data succeeds; this would directly demonstrate that Assumption 1, rather than the RIP or the optimizer, is the load-bearing condition.

Watch

Extended reading notes

Core claim

The central claim is that learning a superoperator of the form (2.1) reduces to low-rank matrix sensing of the Hermitian reshaped matrix $K = R(\mathrm{mat}(K)) = VV^\dagger - UU^\dagger \in \mathbb{C}^{N^2 \times N^2}$, which is not positive semidefinite in general. Under random Pauli measurements, the sensing map satisfies a $(2r, 1/20)$-restricted isometry property once $M \ge C r N^2 \log^6 N$, and then all local minima of the regularized factored objective equal the true $K$ in the noiseless case, with explicit noisy error bounds. The blockwise design instead isolates the first-row blocks $K_{1k}$: with $M_O \ge C r N \log^6 N$ shared Pauli observables and $(3N-2)M_O$ total measurements, the stacked sensing map has the same RIP constant, so joint alternating least squares recovers the entire first row. Algorithm 5 then reconstructs all other block rows from a randomized SVD of the first row and the pseudo-inverse of $J_1^\dagger$, provided the first diagonal block $K^*_{11}$ has rank $r$. Numerical tests show channel and Lindbladian recovery at comparable accuracy, with the blockwise algorithms running roughly 100 to 1000 times faster than the direct random design at the sizes tested.

Load-bearing premise

The whole deterministic completion step presupposes that the first diagonal block of the reshaped superoperator has full rank $r$; if that block is degenerate, for example a low-rank diagonal operator with zero off-diagonal blocks, estimating the first row cannot determine the rest of the matrix and the method breaks down.

Editorial extensions

If this is right

  • If the bounds are right, learning a low-Kraus-rank channel or a low-rank Lindbladian requires $O(r N^2 \log^6 N)$ measurements instead of the $O(N^4)$ of full process tomography.
  • The blockwise first-row design keeps the same measurement scaling as the random design but reduces each ALS subproblem to size $N \times N$, enabling parallel execution and cutting runtime by orders of magnitude (roughly 100 times at $N=8$ and 1000 times at $N=16$ in the reported experiments).
  • Because the recovery guarantees rest on the restricted isometry property rather than positive semidefiniteness, the same algorithm applies to quantum channels ($r_- = 0$) and Lindbladians ($r_- = 1$) with no separate treatment.
  • In the noisy case the error bounds scale polynomially in the noise level and system dimension, and numerical experiments confirm that the Frobenius error decays as the number of observables grows.
  • The first-row assumption can be relaxed: any row whose diagonal block has rank $r$ works, and randomly chosen blocks can in principle be handled through group synchronization.

Reading between the lines

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

  • An implication the paper only gestures at is that the Lindbladian half of the unified claim is not yet an end-to-end experimental protocol, because the paper explicitly sets aside the derivative-estimation step that would supply noisy time-derivative data from actual trajectories.
  • The deterministic completion step inverts the first diagonal block, so Assumption 1 is also a structural prior on the superoperator; one could test how often structured physical noise models, such as diagonal dephasing or sparse jump operators, violate it, since the Haar-random justification covers generic but not structured operators.
  • The same blockwise plus matrix-completion pipeline could be applied to other non-Hermitian low-rank linear maps whenever the measurement model gives access to subblocks; the paper's extension to pseudo-unitary group synchronization points in that direction.
  • A concrete next experiment would run the subset-based joint ALS on data from a real quantum processor with a known Lindbladian noise model, to see whether the measurement-count and runtime advantages survive experimental state-preparation and measurement errors.
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.

Referee Report

5 major / 5 minor

Summary. The paper proposes a unified low-rank matrix sensing framework for learning quantum superoperators, covering both quantum channels (positive-semidefinite Choi-type reshaped matrices) and Lindbladians (indefinite reshaped matrices). The main contributions are: (i) a random measurement design based on tensor products of Pauli operators, with recovery guarantees derived from the restricted isometry property (RIP) and nonconvex landscape results; (ii) a first-row blockwise measurement design that reduces the full N^2 x N^2 recovery problem to N x N sub-block recovery, combined with parallel or joint ALS algorithms and a deterministic reconstruction step from the first row; and (iii) numerical experiments demonstrating large speedups over full tomography and the random design. The stated sample complexity is M = O(r N^2 log^6 N) measurements for both designs, and the blockwise design is claimed to recover all blocks with the same order of measurements.

Significance. If the guarantees were fully correct, the blockwise design would be a valuable practical contribution to low-rank quantum process tomography and Lindbladian learning, offering order-of-magnitude computational savings relative to full N^2 x N^2 matrix sensing. The paper explicitly compares against the CPT-fit method of [41] and shows competitive recovery rates at a fraction of the runtime. The numerical study is fairly extensive, covering channels and Lindbladians at several system sizes. However, several load-bearing technical gaps affect the printed guarantees: the measurement identity used for the random design is incorrect as stated, the RIP rank parameters in the corollaries do not match the requirements of the cited theorems, the implemented algorithm minimizes an objective different from the one whose landscape is analyzed, and the deterministic block reconstruction relies on a rank assumption that fails for physically common channels. These issues are local and fixable in principle, but they must be addressed before the theoretical claims can be accepted.

major comments (5)
  1. [Section 2.1, Proposition 2.7 and Section 3.1] The identity (2.4) is false for complex Hermitian ρ0. The proof uses vec(O)vec(ρ0)† = R(ρ0⊗O), but Lemma 2.2 gives R(B⊗C)=vec(C)vec(B)⊤, so R(ρ0⊗O)=vec(O)vec(ρ0)⊤. The correct relation is ⟨ρ1,O⟩ = ⟨K, \bar ρ0⊗O⟩. A concrete counterexample is the identity channel with ρ0=O=σ_y: the left-hand side of (2.4) is 2, while the right-hand side is -2. Consequently, the random measurement model in Section 3.1, which sets b⋆m0 = tr[(ρm0 ⊗ Om)†K⋆] using (2.4), is inconsistent with the physical measurement tr[O_m^† K(ρ_m0)]. The sensing matrices should be Am = ρm0^⊤ ⊗ Om (or \bar ρm0 ⊗ Om) rather than ρm0 ⊗ Om. Since the transpose of a Pauli matrix is again a Pauli matrix, the RIP and sample-complexity claims O(rN^2 log^6 N) would survive this correction, but Corollary 3.1 as stated is not valid.
  2. [Corollaries 3.1, 3.4, 3.7] The corollaries state that the sensing operator satisfies the (r,δ)-RIP with δ ≤ 1/20 and then invoke Theorems 2.1 and 2.2, which require the (2r,1/20)-RIP. An (r,δ)-RIP bound does not imply a (2r,δ)-RIP bound, so the cited theorems do not apply as written. The fix is local: the corollaries should be restated with the (2r,1/20)-RIP. The required sample size from Theorem 2.4 then becomes M ≥ C(2r)d log^6 d = O(rd log^6 d), which is the same order as claimed. This is a technical correction that does not change the scaling conclusions.
  3. [Section 2.5, Remark 2.14 and Algorithms 1-4] The theoretical landscape guarantees (Theorems 2.1 and 2.2) apply to the regularized objective (2.7) with λ = c/8, but the implemented ALS algorithms, including the Nesterov-accelerated version used in Algorithms 3 and 4, minimize the unregularized objective (2.6)/(Q0). Remark 2.14 explicitly acknowledges this omission and cites empirical evidence, but no analysis of the unregularized landscape is provided. As a result, the numerical recoveries are not covered by the paper's theoretical guarantees. The authors should either implement the regularized objective or prove that the unregularized landscape has the same favorable properties.
  4. [Section 3.4-3.5, Assumption 1 and Algorithm 5] The deterministic reconstruction from the first row requires Assumption 1 (rank K*_11 = r) in order for (3.6) to invert J1^† and recover all rows. This assumption fails for physically standard low-rank channels. For example, the completely dephasing channel Φ(ρ) = Σ_i |i⟩⟨i| ρ |i⟩⟨i| has Kraus rank r=N and its reshaped Choi matrix is diagonal with N ones, so every diagonal block K*_kk has rank 1 < r. In this case the first-row measurements carry no information about the off-diagonal blocks, and Algorithm 5 cannot recover the full matrix. The paper does state Assumption 1 and Remark 3.11 notes the need for some full-rank diagonal block, but the abstract and introduction do not qualify the blockwise guarantee. The main claims should be restated as conditional on this assumption, or the method should be augmented with a row-selection procedure that is guaranteed to find a full-rank diagonal block (which may not exist for certain channels).
  5. [Remark 2.10] The paper asserts that Theorems 2.1 and 2.2, originally proved for real-valued matrices, extend to the complex case with the same proof, but no details or reference are given. Since the quantum sensing problem is intrinsically complex, this extension is load-bearing for all theoretical guarantees. The authors should either provide a proof (e.g., via Wirtinger calculus) or cite a published complex version of the landscape theorems.
minor comments (5)
  1. [Lemma 2.2] There is a typo in the statement of the involution property: 'R((R(A))' should read 'R(R(A))'.
  2. [Section 3.2] The line '⟨ρ⊤0 ⊗ O†, K†⟩ = ⟨Ekl ⊗ O, K⟩' omits the adjoint on O; it is only correct when O is Hermitian. Please clarify the notation and the assumptions in that derivation.
  3. [Section 2, Eq. (2.1)] The text says 'r+, r− > 0', but for quantum channels r− = 0. The condition should be r− ≥ 0.
  4. [Remark 3.12] The performance guarantee for Algorithm 5 is stated in terms of the condition number of J1 and the spectral gap of K1, but these quantities are not bounded under any stated assumptions. The authors should either quantify them or state clearly that the deterministic reconstruction has no finite-sample guarantee.
  5. [Section 4] The numerical experiments use random density matrices and random Hermitian observables from QuTiP, which are not the incoherent Pauli measurements treated in the theoretical sections. The paper acknowledges this in remarks, but the theory-numerics gap should be made explicit in the main text.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the recovery guarantees rest on external RIP and landscape theorems, and the blockwise reconstruction is a stated conditional identity rather than a hidden fit.

full rationale

The paper's derivation chain is self-contained with respect to its cited tools. The random measurement design invokes the Pauli-measurement RIP of Liu ([51]) and the nonconvex landscape results of Ge, Jin, and Zheng ([25]); both are external, parameter-free results with stated assumptions that do not include the target recovery theorem. The blockwise design rewrites the sensing identity via Proposition 2.7 and then solves independent or joint low-rank matrix sensing problems for the first-row blocks, with Corollaries 3.4 and 3.7 following directly from Theorem 2.4 and the external RIP/landscape theorems. The deterministic reconstruction from the first row (Section 3.5) is an explicit algebraic procedure requiring Assumption 1 (rank K*_11 = r); this is an openly stated scope condition, and the paper acknowledges in Remark 3.12 that the resulting error bound depends on unquantified condition numbers and spectral gaps. That is a correctness or applicability limitation, not circularity. The numerical experiments use randomly generated ground-truth superoperators and report empirical recovery rates; no fitted parameter is renamed as a prediction, and the empirical slope reported in Figure 3 is a post hoc observation, not an input to the theoretical claims. The only self-citation, [44] in the related-work discussion of graph dynamical systems, is not load-bearing for any central result. Thus the paper does not reduce its predictions to its inputs by construction.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The central claim rests on external RIP and landscape theorems and on one structural prior (rank of the first diagonal block). No free parameters are fitted to data; the unspecified constants in the sample complexity are standard. No new physical entities are introduced.

assumptions (6)
  • standard math Random Pauli measurements satisfy the RIP with M ≥ C K^2 r d log^6 d (Liu 2011, Theorem 2.3)
    Invoked in Theorem 2.4 and Corollaries 3.1, 3.4, 3.7 as the sample-complexity backbone. The paper adapts the constants but does not reprove it.
  • ad hoc to paper The landscape theorems of Ge, Jin, and Zheng (2017) extend verbatim from real to complex matrices
    Section 2.3 states 'We argue that the same proof holds for the complex case' without giving the argument. The central recovery bounds depend on these theorems.
  • ad hoc to paper The unregularized ALS objective behaves like the regularized objective (2.7) with λ=c/8
    Remark 2.14 says the regularization is omitted and relies on empirical evidence; all landscape guarantees formally apply only to the regularized problem.
  • domain assumption Assumption 1: the first diagonal block K*_11 of the reshaped operator has rank r
    Required for the pseudo-inverse step (3.6) and the full reconstruction (3.7). The paper proves it only for Haar-random unitary eigenbases and notes it fails for diagonal low-rank matrices.
  • domain assumption Noisy estimates of the Lindbladian output, i.e., time derivatives, are available as measurements
    Section 1.1.3 and Section 4.1 explicitly restrict the scope to the second step: learning L from derivative data. The derivative-estimation problem is deferred.
  • ad hoc to paper The (r,δ)-RIP with δ ≤ 1/20 implies the (2r,1/20)-RIP required by Theorems 2.1 and 2.2
    Not proven in the paper and not generally true for the stated sample size; Corollaries 3.1, 3.4, and 3.7 use this implication implicitly.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Unified Blockwise Measurement Design for Learning Quantum Channels and Lindbladians via Low-Rank Matrix Sensing." pith.science (2026). https://pith.science/paper/45JO2EO5

@misc{pith2026250114080,
  author       = {Pith},
  title        = {Pith review of: A Unified Blockwise Measurement Design for Learning Quantum Channels and Lindbladians via Low-Rank Matrix Sensing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/45JO2EO5}},
  note         = {Machine review of arXiv:2501.14080}
}
read the original abstract

Quantum superoperator learning is a pivotal task in quantum information science, enabling accurate reconstruction of unknown quantum operations from measurement data. We propose a robust approach based on the matrix sensing techniques for quantum superoperator learning that extends beyond the positive semidefinite case, encompassing both quantum channels and Lindbladians. We first introduce a randomized measurement design using a near-optimal number of measurements. By leveraging the restricted isometry property (RIP), we provide theoretical guarantees for the identifiability and recovery of low-rank superoperators in the presence of noise. Additionally, we propose a blockwise measurement design that restricts the tomography to the sub-blocks, significantly enhancing performance while maintaining a comparable scale of measurements. We also provide a performance guarantee for this setup. Our approach employs alternating least squares (ALS) with acceleration for optimization in matrix sensing. Numerical experiments validate the efficiency and scalability of the proposed methods.

Figures

Figures reproduced from arXiv: 2501.14080 by the authors.

Figure 1
Figure 1. The recovery rate for different numbers of measurements in channel learning with [PITH_FULL_IMAGE:figures/full_fig_p021_1.png] view at source ↗
Figure 2
Figure 2. The recovery rate and computation time for channel and Lindbladian learning with [PITH_FULL_IMAGE:figures/full_fig_p021_2.png] view at source ↗
Figure 3
Figure 3. The change of Frobenius error, computation time, and number of iterations in ALS, with the [PITH_FULL_IMAGE:figures/full_fig_p022_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Recovery rate with changing N and r. The growth of MO is roughly linear in N and r, so the number of measurements M scales roughly quadratically in N and linearly in r. problem into learning K11 = UV and actually introduced more variables when r ≥ N/2. This redundancy …

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Absent, Not Faint: Fisher-Information Limits and a Logarithmic Measurement-Design Cure for Passive Characterization of Coherent Qubit Noise

    quant-ph 2026-07 conditional novelty 6.0 of 10

    Coherent over-rotation angles are Fisher-null (CRB infinite) in a single fixed-basis histogram at zero angle; adding ⌈log2(n+1)⌉ extra product-Pauli settings restores identifiability, with conditioning, not coverage, ...

Reference graph

Works this paper leans on

81 extracted references · 73 canonical work pages · cited by 1 Pith paper

  1. [41]

    Guaranteed recovery of quantum pro- cesses from few measurements

    Martin Kliesch, Richard Kueng, Jens Eisert, and David Gross. “Guaranteed recovery of quantum pro- cesses from few measurements”. In: Quantum 3 (12, 2019), p. 171. (Visited on 02/27/2024)

    work page 2019

  2. [1]

    Learning a local Hamiltonian from local measure- ments

    Eyal Bairey, Itai Arad, and Netanel H. Lindner. “Learning a local Hamiltonian from local measure- ments”. In: Physical Review Letters 122.2 (2019), p. 020504

    2019

  3. [2]

    Learning the dynamics of open quantum systems from their steady states

    Eyal Bairey, Chu Guo, Dario Poletti, Netanel H. Lindner, and Itai Arad. “Learning the dynamics of open quantum systems from their steady states”. In: New Journal of Physics 22.3 (2020), p. 032001. 23

    work page 2020

  4. [3]

    Quantum process tomography of unitary and near-unitary Maps

    Charles H. Baldwin, Amir Kalev, and Ivan H. Deutsch. “Quantum process tomography of unitary and near-unitary Maps”. In: Physical Review A 90.1 (17, 2014), p. 012110

    work page 2014

  5. [4]

    On the complexity of approximating the diamond norm

    Avraham Ben-Aroya and Amnon Ta-Shma. “On the complexity of approximating the diamond norm”. In: arXiv preprint arXiv:0902.3397 (2009)

    work page Pith review arXiv 2009

  6. [5]

    Iterative hard thresholding for compressed sensing

    Thomas Blumensath and Mike E. Davies. “Iterative hard thresholding for compressed sensing”. In: Applied and computational harmonic analysis 27.3 (2009), pp. 265–274

    work page 2009

  7. [6]

    Normalized iterative hard thresholding: Guaranteed stability and performance

    Thomas Blumensath and Mike E. Davies. “Normalized iterative hard thresholding: Guaranteed stability and performance”. In: IEEE Journal of selected topics in signal processing 4.2 (2010), pp. 298–309

    work page 2010

  8. [7]

    Parameter estimation of ODE’s via nonparametric estimators

    Nicolas J-B. Brunel. “Parameter estimation of ODE’s via nonparametric estimators”. In: Electronic Journal of Statistics 2 (2008), pp. 1242–1267

    work page 2008

Show all 81 references
  1. [8]

    A tracking approach to parameter estimation in linear ordinary differential equations

    Nicolas J-B. Brunel and Quentin Clairon. “A tracking approach to parameter estimation in linear ordinary differential equations”. In: Electronic Journal of Statistics 9.2 (2015), pp. 2903–2949

    2015

  2. [9]

    Random quantum operations

    Wojciech Bruzda, Valerio Cappellini, Hans-J¨ urgen Sommers, and Karol˙Zyczkowski. “Random quantum operations”. In: Physics Letters A 373.3 (2009), pp. 320–324

    2009

  3. [10]

    A singular value thresholding algorithm for matrix completion

    Jian-Feng Cai, Emmanuel J. Cand` es, and Zuowei Shen. “A singular value thresholding algorithm for matrix completion”. In: SIAM Journal on optimization 20.4 (2010), pp. 1956–1982

    2010

  4. [11]

    The power of convex relaxation: Near-optimal matrix comple- tion

    Emmanuel J. Cand` es and Terence Tao. “The power of convex relaxation: Near-optimal matrix comple- tion”. In: IEEE transactions on information theory 56.5 (2010), pp. 2053–2080

    2010

  5. [12]

    Exact matrix completion via convex optimization

    Emmanuel J. Cand` es and Benjamin Recht. “Exact matrix completion via convex optimization”. In: Communications of the ACM 55.6 (2012), pp. 111–119

    2012

  6. [13]

    Stable signal recovery from incomplete and inaccurate measurements

    Emmanuel J. Cand` es, Justin K Romberg, and Terence Tao. “Stable signal recovery from incomplete and inaccurate measurements”. In: Communications on Pure and Applied Mathematics: A Journal Issued by the Courant Institute of Mathematical Sciences 59.8 (2006), pp. 1207–1223

    2006

  7. [14]

    Learning Quantum Hamiltonians from Single-Qubit Measurements

    Liangyu Che, Chao Wei, Yulei Huang, Dafa Zhao, Shunzhong Xue, Xinfang Nie, Jun Li, Dawei Lu, and Tao Xin. “Learning Quantum Hamiltonians from Single-Qubit Measurements”. In:Physical Review Research 3.2 (2021), p. 023246

    2021

  8. [15]

    Completely positive linear maps on complex matrices

    Man-Duen Choi. “Completely positive linear maps on complex matrices”. In: Linear algebra and its applications 10.3 (1975), pp. 285–290

    1975

  9. [16]

    Prescription for experimental determination of the dynamics of a quantum black box

    Isaac L. Chuang and Michael A. Nielsen. “Prescription for experimental determination of the dynamics of a quantum black box”. In: Journal of Modern Optics 44.11-12 (1997), pp. 2455–2467

    1997

  10. [17]

    Iteratively reweighted least squares minimization for sparse recovery

    Ingrid Daubechies, Ronald DeVore, Massimo Fornasier, and C. Sinan G¨ unt¨ urk. “Iteratively reweighted least squares minimization for sparse recovery”. In:Communications on Pure and Applied Mathematics: A Journal Issued by the Courant Institute of Mathematical Sciences 63.1 (2...

    2010

  11. [18]

    Davenport, Marco F

    Mark A. Davenport, Marco F. Duarte, Yonina C. Eldar, and Gitta Kutyniok. Introduction to compressed sensing. 2012

    2012

  12. [19]

    An overview of low-rank matrix recovery from incomplete observations

    Mark A. Davenport and Justin Romberg. “An overview of low-rank matrix recovery from incomplete observations”. In: IEEE Journal of Selected Topics in Signal Processing 10.4 (2016), pp. 608–622

    2016

  13. [20]

    The rotation of eigenvectors by a perturbation. III

    Chandler Davis and William Morton Kahan. “The rotation of eigenvectors by a perturbation. III”. In: SIAM Journal on Numerical Analysis 7.1 (1970), pp. 1–46

    1970

  14. [21]

    On the eigenvalues of random matrices

    Persi Diaconis and Mehrdad Shahshahani. “On the eigenvalues of random matrices”. In: Journal of Applied Probability 31.A (1994), pp. 49–62

    1994

  15. [22]

    Compressed-sensing Lindbladian quantum tomography with trapped ions

    Dmitrii Dobrynin, Lorenzo Cardarelli, Markus M¨ uller, and Alejandro Bermudez. “Compressed-sensing Lindbladian quantum tomography with trapped ions”. In: arXiv preprint arXiv:2403.07462 (2024)

    2024 arXiv

  16. [23]

    Compressed sensing

    David L. Donoho. “Compressed sensing”. In: IEEE Transactions on information theory 52.4 (2006), pp. 1289–1306

    2006

  17. [24]

    Quantum tomography via compressed sensing: Error bounds, sample complexity and efficient estimators

    Steven T. Flammia, David Gross, Yi-Kai Liu, and Jens Eisert. “Quantum tomography via compressed sensing: Error bounds, sample complexity and efficient estimators”. In: New Journal of Physics 14.9 (2012), p. 095022

    2012

  18. [25]

    No spurious local minima in nonconvex low rank problems: A unified geometric analysis

    Rong Ge, Chi Jin, and Yi Zheng. “No spurious local minima in nonconvex low rank problems: A unified geometric analysis”. In: International Conference on Machine Learning . PMLR. 2017, pp. 1233–1242. 24

    2017

  19. [26]

    Distance measures to compare real and ideal quantum processes

    Alexei Gilchrist, Nathan K. Langford, and Michael A. Nielsen. “Distance measures to compare real and ideal quantum processes”. In: Physical Review A—Atomic, Molecular, and Optical Physics 71.6 (2005), p. 062310

    2005

  20. [27]

    Completely positive dynamical semigroups of N-level systems

    Vittorio Gorini, Andrzej Kossakowski, and Ennackal Chandy George Sudarshan. “Completely positive dynamical semigroups of N-level systems”. In: Journal of Mathematical Physics 17.5 (1976), pp. 821– 825

    1976

  21. [28]

    Robust online Hamil- tonian learning

    Christopher E. Granade, Christopher Ferrie, Nathan Wiebe, and David G. Cory. “Robust online Hamil- tonian learning”. In: New Journal of Physics 14.10 (2012), p. 103013

    2012

  22. [29]

    Recovering low-rank matrices from few coefficients in any basis

    David Gross. “Recovering low-rank matrices from few coefficients in any basis”. In: IEEE Transactions on Information Theory 57.3 (2011), pp. 1548–1566

    2011

  23. [30]

    Quantum state to- mography via compressed sensing

    David Gross, Yi-Kai Liu, Steven T. Flammia, Stephen Becker, and Jens Eisert. “Quantum state to- mography via compressed sensing”. In: Physical review letters 105.15 (2010), p. 150401

    2010

  24. [31]

    Practical black box Hamiltonian learning

    Anran Gu, Lukasz Cincio, and Patrick J. Coles. “Practical black box Hamiltonian learning”. In: arXiv preprint arXiv:2206.15464 (2022)

    2022 arXiv

  25. [32]

    On a combination of alternating minimization and Nesterov’s momentum

    Sergey Guminov, Pavel Dvurechensky, Nazarii Tupitsa, and Alexander Gasnikov. “On a combination of alternating minimization and Nesterov’s momentum”. In:International conference on machine learning. PMLR. 2021, pp. 3886–3898

    2021

  26. [33]

    Finding structure with randomness: Prob- abilistic algorithms for constructing approximate matrix decompositions

    Nathan Halko, Per-Gunnar Martinsson, and Joel A. Tropp. “Finding structure with randomness: Prob- abilistic algorithms for constructing approximate matrix decompositions”. In: SIAM review 53.2 (2011), pp. 217–288

    2011

  27. [34]

    Forcing function diagnostics for nonlinear dynamics

    Giles Hooker. “Forcing function diagnostics for nonlinear dynamics”. In: Biometrics 65.3 (2009), pp. 928– 936

    2009

  28. [35]

    Quantum process tomography and Lindblad estimation of a solid-state qubit

    Mark Howard, Jason Twamley, Christoffer Wittmann, Torsten Gaebel, Fedor Jelezko, and J¨ org Wrachtrup. “Quantum process tomography and Lindblad estimation of a solid-state qubit”. In: New Journal of Physics 8.3 (2006), p. 33

    2006

  29. [36]

    Guaranteed rank minimization via singular value projection

    Prateek Jain, Raghu Meka, and Inderjit Dhillon. “Guaranteed rank minimization via singular value projection”. In: Advances in Neural Information Processing Systems 23 (2010)

    2010

  30. [37]

    Low-rank matrix completion using alternating minimization

    Prateek Jain, Praneeth Netrapalli, and Sujay Sanghavi. “Low-rank matrix completion using alternating minimization”. In: Proceedings of the forty-fifth annual ACM symposium on Theory of computing. 2013, pp. 665–674

    2013

  31. [38]

    QuTiP: An open-source Python framework for the dynamics of open quantum systems

    J. Robert Johansson, Paul D. Nation, and Franco Nori. “QuTiP: An open-source Python framework for the dynamics of open quantum systems”. In: Computer physics communications 183.8 (2012), pp. 1760– 1772

    2012

  32. [39]

    Algorithms for nonnegative matrix and tensor factoriza- tions: A unified view based on block coordinate descent framework

    Jingu Kim, Yunlong He, and Haesun Park. “Algorithms for nonnegative matrix and tensor factoriza- tions: A unified view based on block coordinate descent framework”. In: Journal of Global Optimization 58 (2014), pp. 285–319

    2014

  33. [40]

    Phase retrieval using unitary 2-designs

    Shelby Kimmel and Yi-Kai Liu. “Phase retrieval using unitary 2-designs”. In: 2017 International Con- ference on Sampling Theory and Applications (SampTA) . IEEE. 2017, pp. 345–349

    2017

  34. [42]

    Principal component analysis of three-mode data by means of alternating least squares algorithms

    Pieter M. Kroonenberg and Jan De Leeuw. “Principal component analysis of three-mode data by means of alternating least squares algorithms”. In: Psychometrika 45 (1980), pp. 69–97

    1980

  35. [43]

    Low rank matrix recovery from rank one mea- surements

    Richard Kueng, Holger Rauhut, and Ulrich Terstiege. “Low rank matrix recovery from rank one mea- surements”. In: Applied and Computational Harmonic Analysis 42.1 (2017), pp. 88–116

    2017

  36. [44]

    Interacting particle systems on networks: Joint inference of the network and the interaction kernel

    Quanjun Lang, Xiong Wang, Fei Lu, and Mauro Maggioni. “Interacting particle systems on networks: Joint inference of the network and the interaction kernel”. In: arXiv preprint arXiv:2402.08412 (2024)

    2024

  37. [45]

    Randomly initialized alternating least squares: Fast convergence for matrix sensing

    Kiryung Lee and Dominik St¨ oger. “Randomly initialized alternating least squares: Fast convergence for matrix sensing”. In: SIAM Journal on Mathematics of Data Science 5.3 (2023), pp. 774–799

    2023

  38. [46]

    Parameter estimation of ordinary differential equations

    Zhengfeng Li, Michael R. Osborne, and Tania Prvan. “Parameter estimation of ordinary differential equations”. In: IMA Journal of Numerical Analysis 25.2 (2005), pp. 264–285. 25

    2005

  39. [47]

    Hamiltonian tomography via quantum quench

    Zhi Li, Liujun Zou, and Timothy H Hsieh. “Hamiltonian tomography via quantum quench”. In: Physical review letters 124.16 (2020), p. 160502

    2020

  40. [48]

    Adaptive geometric multiscale approximations for intrinsically low-dimensional data

    Wenjing Liao and Mauro Maggioni. “Adaptive geometric multiscale approximations for intrinsically low-dimensional data”. In: Journal of machine learning research 20.98 (2019), pp. 1–63

    2019

  41. [49]

    On the generators of quantum dynamical semigroups

    Goran Lindblad. “On the generators of quantum dynamical semigroups”. In: Communications in math- ematical physics 48 (1976), pp. 119–130

    1976

  42. [50]

    A unified approach to synchronization problems over subgroups of the orthogonal group

    Huikang Liu, Man-Chung Yue, and Anthony Man-Cho So. “A unified approach to synchronization problems over subgroups of the orthogonal group”. In: Applied and Computational Harmonic Analysis 66 (2023), pp. 320–372

    2023

  43. [51]

    Universal low-rank matrix recovery from Pauli measurements

    Yi-Kai Liu. “Universal low-rank matrix recovery from Pauli measurements”. In: Advances in Neural Information Processing Systems 24 (2011)

    2011

  44. [52]

    Learning k-body Hamiltonians via compressed sensing

    Muzhou Ma, Steven T. Flammia, John Preskill, and Yu Tong. “Learning k-body Hamiltonians via compressed sensing”. In: arXiv preprint arXiv:2410.18928 (2024)

    2024 arXiv

  45. [53]

    Dynamic anomalography: Tracking network anomalies via sparsity and low rank

    Morteza Mardani, Gonzalo Mateos, and Georgios B. Giannakis. “Dynamic anomalography: Tracking network anomalies via sparsity and low rank”. In: IEEE Journal of Selected Topics in Signal Processing 7.1 (2012), pp. 50–66

    2012

  46. [54]

    A randomized algorithm for the decom- position of matrices

    Per-Gunnar Martinsson, Vladimir Rokhlin, and Mark Tygert. “A randomized algorithm for the decom- position of matrices”. In: Applied and Computational Harmonic Analysis 30.1 (2011), pp. 47–68

    2011

  47. [55]

    Nesterov acceleration of alternating least squares for canonical tensor decomposition: Momentum step size selection and restart mechanisms

    Drew Mitchell, Nan Ye, and Hans De Sterck. “Nesterov acceleration of alternating least squares for canonical tensor decomposition: Momentum step size selection and restart mechanisms”. In: Numerical Linear Algebra with Applications 27.4 (2020), e2297

    2020

  48. [56]

    Quantum-process tomography: Resource analysis of different strategies

    Masoud Mohseni, Ali T. Rezakhani, and Daniel A. Lidar. “Quantum-process tomography: Resource analysis of different strategies”. In: Physical Review A—Atomic, Molecular, and Optical Physics 77.3 (2008), p. 032322

    2008

  49. [57]

    Self-adjoint elements in the pseudo-unitary group U (p, p)

    Sachin Munshi and Rongwei Yang. “Self-adjoint elements in the pseudo-unitary group U (p, p)”. In: Linear Algebra and its Applications 560 (2019), pp. 100–113

    2019

  50. [58]

    Joint covariate selection and joint subspace selection for multiple classification problems

    Guillaume Obozinski, Ben Taskar, and Michael I. Jordan. “Joint covariate selection and joint subspace selection for multiple classification problems”. In: Statistics and Computing 20 (2010), pp. 231–252

    2010

  51. [59]

    Fitting quantum noise models to tomography data

    Emilio Onorati, Tamara Kohler, and Toby S. Cubitt. “Fitting quantum noise models to tomography data”. In: Quantum 7 (2023), p. 1197

    2023

  52. [60]

    Non-square matrix sensing without spurious local minima via the Burer-Monteiro approach

    Dohyung Park, Anastasios Kyrillidis, Constantine Carmanis, and Sujay Sanghavi. “Non-square matrix sensing without spurious local minima via the Burer-Monteiro approach”. In: Artificial Intelligence and Statistics. PMLR. 2017, pp. 65–74

    2017

  53. [61]

    Block acceleration without momentum: On optimal stepsizes of block gradient descent for least-squares

    Liangzu Peng and Wotao Yin. “Block acceleration without momentum: On optimal stepsizes of block gradient descent for least-squares”. In: arXiv preprint arXiv:2405.16020 (2024)

    2024 arXiv

  54. [62]

    Guaranteed minimum-rank solutions linear matrix equations via nuclear norm minimization

    Benjamin Recht, Maryam Fazel, and Pablo A. Parrilo. “Guaranteed minimum-rank solutions linear matrix equations via nuclear norm minimization”. In: SIAM Review 52.3 (2010), pp. 471–501

    2010

  55. [63]

    Compressed sensing quantum process tomography for superconducting quantum gates

    Andrey V. Rodionov, Andrzej Veitia, R. Barends, J. Kelly, Daniel Sank, J. Wenner, John M. Martinis, Robert L. Kosut, and Alexander N. Korotkov. “Compressed sensing quantum process tomography for superconducting quantum gates”. In: Physical Review B 90.14 (2014), p. 144504

    2014

  56. [64]

    Lindblad tomography of a superconducting quantum processor

    Gabriel O. Samach, Ami Greene, Johannes Borregaard, Matthias Christandl, Joseph Barreto, David K. Kim, Christopher M. McNally, Alexander Melville, Bethany M. Niedzielski, Youngkyu Sung, et al. “Lindblad tomography of a superconducting quantum processor”. In: Physical Review Ap...

    2022

  57. [65]

    Efficient measurement of quantum dynamics via compressive sensing

    A. Shabani, R.L. Kosut, M. Mohseni, H. Rabitz, M.A. Broome, M.P. Almeida, A. Fedrizzi, and A.G. White. “Efficient measurement of quantum dynamics via compressive sensing”. In: Physical review letters 106.10 (2011), p. 100401

    2011

  58. [66]

    da Silva

    Marcus P. da Silva. matlab-diamond-norm. 2014

    2014

  59. [67]

    Angular synchronization by eigenvectors and semidefinite programming

    Amit Singer. “Angular synchronization by eigenvectors and semidefinite programming”. In: Applied and computational harmonic analysis 30.1 (2011), pp. 20–36. 26

    2011

  60. [68]

    Efficient and robust estimation of many-qubit Hamiltonians

    Daniel Stilck Fran¸ ca, Liubov A. Markovich, V.V. Dobrovitski, Albert H. Werner, and Johannes Borre- gaard. “Efficient and robust estimation of many-qubit Hamiltonians”. In: Nature Communications 15.1 (2024), p. 311

    2024

  61. [69]

    Convergence of a block coordinate descent method for nondifferentiable minimization

    Paul Tseng. “Convergence of a block coordinate descent method for nondifferentiable minimization”. In: Journal of optimization theory and applications 109 (2001), pp. 475–494

    2001

  62. [70]

    Approximation with Kronecker products

    Charles F. Van Loan and Nikos Pitsianis. “Approximation with Kronecker products”. In: Linear Algebra for Large Scale and Real-Time Applications . Dordrecht: Springer Netherlands, 1993, pp. 293–314

    1993

  63. [71]

    Simulation-assisted learning of open quantum systems

    Ke Wang and Xiantao Li. “Simulation-assisted learning of open quantum systems”. In: Quantum 8 (2024), p. 1407

    2024

  64. [72]

    Exact and stable recovery of rotations for robust synchronization

    Lanhui Wang and Amit Singer. “Exact and stable recovery of rotations for robust synchronization”. In: Information and Inference: A Journal of the IMA 2.2 (2013), pp. 145–193

    2013

  65. [73]

    Semidefinite programs for completely bounded norms

    John Watrous. “Semidefinite programs for completely bounded norms”. In: arXiv preprint arXiv:0901.4709 (2009)

    2009 arXiv

  66. [74]

    Simpler semidefinite programs for completely bounded norms

    John Watrous. “Simpler semidefinite programs for completely bounded norms”. In: arXiv preprint arXiv:1207.5726 (2012)

    2012 arXiv

  67. [75]

    The power of preconditioning in overparameterized low-rank matrix sensing

    Xingyu Xu, Yandi Shen, Yuejie Chi, and Cong Ma. “The power of preconditioning in overparameterized low-rank matrix sensing”. In: International Conference on Machine Learning. PMLR. 2023, pp. 38611– 38654

    2023

  68. [76]

    A block coordinate descent method for regularized multiconvex opti- mization with applications to nonnegative tensor factorization and completion

    Yangyang Xu and Wotao Yin. “A block coordinate descent method for regularized multiconvex opti- mization with applications to nonnegative tensor factorization and completion”. In: SIAM Journal on imaging sciences 6.3 (2013), pp. 1758–1789

    2013

  69. [77]

    On global convergence of alternating least squares for tensor approximation

    Yuning Yang. “On global convergence of alternating least squares for tensor approximation”. In: Com- putational Optimization and Applications 84.2 (2023), pp. 509–529

    2023

  70. [78]

    Practical and efficient Hamiltonian learning

    Wenjun Yu, Jie Sun, Zheng Han, and Xiao Yuan. “Practical and efficient Hamiltonian learning”. In: arXiv preprint arXiv:2201.00190 (2022)

    2022 arXiv

  71. [79]

    Convergence analysis for rectangular matrix completion using Burer- Monteiro factorization and gradient descent

    Qinqing Zheng and John Lafferty. “Convergence analysis for rectangular matrix completion using Burer- Monteiro factorization and gradient descent”. In: arXiv preprint arXiv:1605.07051 (2016)

    2016 arXiv

  72. [80]

    Efficient matrix sensing using rank-1 Gaussian mea- surements

    Kai Zhong, Prateek Jain, and Inderjit Dhillon. “Efficient matrix sensing using rank-1 Gaussian mea- surements”. In: Algorithmic Learning Theory: 26th International Conference, ALT 2015, Banff, AB, Canada, October 4-6, 2015, Proceedings 26 . Springer. 2015, pp. 3–18

    2015

  73. [81]

    Optimal short-time measurements for Hamiltonian learning

    Alon Zubida, Eyal Yitzhaki, Netanel H. Lindner, and Eyal Bairey. “Optimal short-time measurements for Hamiltonian learning”. In: arXiv preprint arXiv:2108.08824 (2021). 27

    2021 arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.