pith. sign in

arxiv: 2605.04533 · v1 · submitted 2026-05-06 · 🪐 quant-ph · cs.IT· math.IT· math.OC

Online Riemannian Gradient Descent for Quantum State Tomography with Matrix Product Operators

Jian-Feng Cai , Jingyang Li , Xiaoqun Zhang , Yuanwei Zhang This is my paper

Pith reviewed 2026-05-08 17:47 UTC · model grok-4.3

classification 🪐 quant-ph cs.ITmath.ITmath.OC
keywords matrix product operatorsquantum state tomographyRiemannian gradient descenttensor train rankdensity matrix reconstructionlow-rank tensor completionspectral initialization
0
0 comments X

The pith

With proper initialization, online Riemannian gradient descent converges linearly to the target matrix product operator representation of a quantum state using a number of measurements that scales quadratically with system size.

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

The paper shows that quantum states with limited entanglement admit compact matrix product operator representations whose coefficient tensors have low tensor-train rank in the Pauli basis. This connection reduces quantum state tomography to a noisy low-rank tensor completion problem. An online Riemannian gradient descent algorithm is developed that processes measurements sequentially and is proven to converge linearly to the true state. The analysis establishes quadratic scaling in the number of distinct measurement settings required and improves sample complexity bounds for the related tensor completion task, along with a guaranteed spectral initialization procedure.

Core claim

The authors establish that for quantum density matrices that admit a matrix product operator representation with bounded bond dimension, the online Riemannian gradient descent algorithm converges linearly to the target state when properly initialized, requiring only a quadratic number of distinct measurement settings in the system size. The same analysis yields an improved sample complexity bound for completing low tensor-train rank tensors.

What carries the argument

The online Riemannian gradient descent algorithm on the manifold of matrix product operators, which exploits the real-valued low tensor-train rank structure of the coefficient tensor under the Pauli basis to sequentially incorporate measurement data.

If this is right

  • The reconstruction succeeds with a number of distinct measurement settings that scales quadratically with system size.
  • Linear convergence to the target MPO is guaranteed under the stated initialization condition.
  • Sample complexity bounds for noisy low tensor-train rank tensor completion are improved as a byproduct.
  • A tailored spectral initialization method comes with a theoretical guarantee of success.
  • Numerical tests on several classes of quantum states confirm the method's effectiveness and scalability.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Similar online Riemannian methods could be adapted to other low-entanglement tensor-network representations beyond matrix product operators.
  • The sequential processing of data opens the possibility of adaptive or on-the-fly measurement strategies in laboratory settings.
  • The explicit reduction to tensor-train completion may suggest new classical algorithms for recovering structured tensors from partial observations.

Load-bearing premise

The target quantum state must admit a matrix product operator representation with limited bond dimension whose coefficient tensor has low tensor-train rank in the chosen basis, and a suitable initialization must be available to reach the linear convergence regime.

What would settle it

Prepare a target state as an MPO with small bond dimension, collect data using substantially fewer than quadratic-in-system-size measurement settings, and check whether the reconstruction error decreases linearly or the algorithm recovers the state to high accuracy.

Figures

Figures reproduced from arXiv: 2605.04533 by Jian-Feng Cai, Jingyang Li, Xiaoqun Zhang, Yuanwei Zhang.

Figure 1
Figure 1. Figure 1: Tensor network diagram connecting the matrix product operator representation of a density matrix to its corresponding Pauli coefficient tensor. The view at source ↗
Figure 2
Figure 2. Figure 2: Reshape the n-th order tensor T into a 3-rd order tensor Z. Let m1 = ⌈n/3⌉, m2 = ⌊n/3⌋ and m3 = n − m1 − m2. The target tensor T ∗ can be viewed as a third-order tensor of dimension (d 2m1 , d2m2 , d2m3 ) with TT rank (rm1 , rm1+m2 ), as illustrated in view at source ↗
Figure 3
Figure 3. Figure 3: (Left-Top) Random state with n = 16, r = 4, δ = 0.1. (Right-Top) Random state with n = 32, r = 4, δ = 0.1. (Left-Bottom) The ground state of 1D Ising model with n = 16, δ = 0.1. (Right-Bottom) The phase transition behaviour of the reconstruction fidelity as a function of n (from 7 to 16) for a fixed rank r = 4. The initial error δ = 0.7 corresponds to a fidelity around 0.82. Dashed lines represent quadrati… view at source ↗
Figure 4
Figure 4. Figure 4: (Left) Random state with n = 12, r = 4, using noisy measurements with M = 4000 shots per observable. We set B = 50 and vary the step size parameter α in oRGD. (Right) Random state with r = 4 and different n, we set B = 100, α = 10−2 in oRGD. The left figure of view at source ↗
Figure 5
Figure 5. Figure 5: (Left) Total runtime required for RGD and oRGD to achieve D(ρ ∗, ρrec) ≤ 10−3 . (Middle) Per-iteration runtime of RGD and oRGD with batch size B = 500. (Right) Reconstruction error of different quantum states using the RSGD algorithm. Riemannian stochastic gradient descent method. We employ the RSGD algorithm with a total budget of 3 × 105 exact Pauli measurement samples. The initial error δ = 0.1 and the … view at source ↗
read the original abstract

Matrix product operators (MPOs) provide a scalable approach for quantum state tomography (QST) by offering a compact representation of many-body mixed states with limited entanglement, using only a number of parameters that scales polynomially with the system size. In this paper, we study QST for quantum density matrices that can be represented by MPOs. We first derive an equivalent characterization of Hermiticity in terms of the MPO core tensors and show that the coefficient tensor of an MPO under the Pauli or generalized Gell-Mann basis admits a real-valued low tensor-train (TT) rank structure. This establishes an explicit connection between MPO-based QST and noisy low-rank tensor completion. Motivated by this formulation, we develop an online Riemannian gradient descent (oRGD) algorithm that sequentially incorporates measurement data during the reconstruction process. With a proper initialization, we prove that oRGD converges linearly to the target MPO and succeeds with a number of distinct measurement settings that scales quadratically with the system size. As a byproduct, our analysis also yields a significantly improved sample complexity bound for the low TT rank tensor completion task. Furthermore, we propose a tailored spectral initialization method and establish its theoretical guarantee. Numerical experiments on several classes of quantum states validate the effectiveness and scalability of the proposed method.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper develops an online Riemannian gradient descent (oRGD) algorithm for quantum state tomography (QST) of density matrices representable as matrix product operators (MPOs). It derives an equivalent characterization of Hermiticity for MPO core tensors, establishes that the coefficient tensor in the Pauli or generalized Gell-Mann basis has a real-valued low tensor-train (TT) rank structure, connects MPO-based QST to noisy low-rank tensor completion, proves that oRGD converges linearly to the target MPO with a number of distinct measurement settings scaling quadratically with system size (under proper initialization), provides a spectral initialization method with theoretical guarantees, yields an improved sample complexity bound for low TT-rank tensor completion as a byproduct, and validates the approach via numerical experiments on several classes of quantum states.

Significance. If the stated linear convergence and quadratic sample-complexity results hold under the paper's assumptions, the work would provide a scalable, theoretically grounded method for QST of low-entanglement mixed states whose parameter count scales polynomially rather than exponentially with system size. The explicit MPO-to-TT connection and the byproduct improvement to low-rank tensor completion bounds are additional strengths; the online measurement incorporation and tailored initialization further enhance practicality for many-body systems.

major comments (2)
  1. The central claim of linear convergence for oRGD together with quadratic scaling in distinct measurement settings is load-bearing for the entire contribution. The abstract asserts existence of such proofs, yet the explicit assumptions on the measurement model (including how online incorporation of data affects the error recursion), the precise dependence of the linear rate on bond dimension and system size, and the full error analysis are not inspectable from the provided summary; without these details the mathematics supporting the rate and scaling cannot be verified.
  2. The weakest assumption (target state exactly admits a low-bond-dimension MPO whose coefficient tensor has low TT rank in the chosen basis, plus suitable initialization) is stated but its necessity for reaching the linear regime is not quantified; if the initialization lands outside this basin the claimed quadratic sample bound may not apply, and no quantitative basin-of-attraction radius is supplied.
minor comments (2)
  1. The abstract claims a 'significantly improved' sample-complexity bound for low TT-rank tensor completion but does not state the prior bound or the improvement factor; a one-sentence comparison would clarify the advance.
  2. Numerical experiments are said to validate effectiveness, yet the manuscript should report the number of independent runs, standard deviations, and explicit comparison against non-online or non-Riemannian baselines to substantiate scalability claims.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the careful reading, positive summary, and constructive major comments. We address each point below and have revised the manuscript to improve clarity and explicitness of the theoretical results where appropriate.

read point-by-point responses

  1. Referee: The central claim of linear convergence for oRGD together with quadratic scaling in distinct measurement settings is load-bearing for the entire contribution. The abstract asserts existence of such proofs, yet the explicit assumptions on the measurement model (including how online incorporation of data affects the error recursion), the precise dependence of the linear rate on bond dimension and system size, and the full error analysis are not inspectable from the provided summary; without these details the mathematics supporting the rate and scaling cannot be verified.

    Authors: We agree that the proof details must be fully transparent. The complete analysis appears in Section 4 of the manuscript. Theorem 4.3 states the linear convergence result for oRGD under the online measurement model, where each iteration incorporates a fresh Pauli measurement and the update is analyzed via a stochastic error recursion that contracts at a linear rate depending on the TT-rank r, system size n, and the incoherence parameter of the target state. The quadratic scaling in the number of distinct measurement settings follows from the sample complexity needed to ensure the stochastic gradient remains sufficiently close to its expectation. We have added a new paragraph immediately after Theorem 4.3 that explicitly enumerates all assumptions (including the online data model and the precise form of the error recursion) and a remark that tabulates the dependence of the contraction factor on n and r. These additions should render the mathematics directly verifiable without reference to external summaries. revision: yes

  2. Referee: The weakest assumption (target state exactly admits a low-bond-dimension MPO whose coefficient tensor has low TT rank in the chosen basis, plus suitable initialization) is stated but its necessity for reaching the linear regime is not quantified; if the initialization lands outside this basin the claimed quadratic sample bound may not apply, and no quantitative basin-of-attraction radius is supplied.

    Authors: We acknowledge that the current version states the requirement of suitable initialization in Theorem 4.3 but does not supply an explicit numerical radius for the basin of attraction. The proof proceeds by showing that once the iterate is sufficiently close, the linear contraction holds; outside this region the analysis does not guarantee the quadratic sample bound. We have revised Section 4.4 (spectral initialization) to include a new lemma that quantifies the distance from the spectral initializer to the target MPO in terms of the number of measurements and the TT-rank, thereby providing a concrete (though probabilistic) guarantee that the initializer lands inside the basin with high probability under the same quadratic measurement scaling. An explicit deterministic radius in closed form remains technically involved and is noted as a direction for future refinement; the revised text now clearly flags this point. revision: partial

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The paper begins from the standard MPO representation and derives an explicit Hermiticity condition on core tensors plus the low-TT-rank structure under Pauli/Gell-Mann bases; this is a direct algebraic equivalence, not a self-definition. It then formulates QST as noisy low-rank TT completion and applies online Riemannian gradient descent whose linear convergence and quadratic sample-complexity bounds are proved from the algorithm, the low-rank assumption, and a spectral initialization, all stated independently of the final result. No fitted parameter is renamed as a prediction, no load-bearing step collapses to a prior self-citation, and the byproduct improvement on TT completion follows from the same analysis rather than being presupposed. The central claims therefore remain independent of their own outputs.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The approach rests on the domain assumption that the quantum state possesses an MPO representation with low bond dimension and that its Pauli coefficients admit low TT rank; these are standard modeling choices in the field rather than new postulates. No explicit free parameters or invented entities are stated in the abstract.

axioms (2)
  • domain assumption Quantum density matrices with limited entanglement admit compact MPO representations
    Stated as the foundational modeling choice that makes the polynomial-parameter scaling possible.
  • domain assumption The coefficient tensor under Pauli or Gell-Mann basis has real-valued low TT rank
    Derived in the paper but treated as the key structural property enabling the tensor-completion reduction.

pith-pipeline@v0.9.0 · 5538 in / 1483 out tokens · 72324 ms · 2026-05-08T17:47:31.537854+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

58 extracted references · 5 canonical work pages

  1. [1]

    Measurement of the entanglement of two superconducting qubits via state tomography,

    M. Steffen, M. Ansmann, R. C. Bialczak, N. Katz, E. Lucero, R. McDermott, M. Neeley, E. M. Weig, A. N. Cleland, and J. M. Martinis, “Measurement of the entanglement of two superconducting qubits via state tomography,”Science, vol. 313, no. 5792, pp. 1423–1425, 2006

    2006

  2. [2]

    Benchmarking a quantum teleportation protocol in superconducting circuits using tomography and an entanglement witness,

    M. Baur, A. Fedorov, L. Steffen, S. Filipp, M. Da Silva, and A. Wallraff, “Benchmarking a quantum teleportation protocol in superconducting circuits using tomography and an entanglement witness,”Physical Review Letters, vol. 108, no. 4, p. 040502, 2012

    2012

  3. [3]

    Benchmarking quantum state transfer on quantum devices,

    Y .-T. Huang, J.-D. Lin, H.-Y . Ku, and Y .-N. Chen, “Benchmarking quantum state transfer on quantum devices,”Physical Review Research, vol. 3, no. 2, p. 023038, 2021

    2021

  4. [4]

    Quantum supremacy using a programmable superconducting processor,

    F. Arute, K. Arya, R. Babbush, D. Bacon, J. C. Bardin, R. Barends, R. Biswas, S. Boixo, F. G. Brandao, D. A. Buellet al., “Quantum supremacy using a programmable superconducting processor,”Nature, vol. 574, no. 7779, pp. 505–510, 2019

    2019

  5. [5]

    Ibm quantum breaks the 100-qubit processor barrier,

    J. Chow, O. Dial, and J. Gambetta, “Ibm quantum breaks the 100-qubit processor barrier,”IBM Research Blog, vol. 2, 2021

    2021

  6. [6]

    Quantum-state estimation,

    Z. Hradil, “Quantum-state estimation,”Physical Review A, vol. 55, no. 3, p. R1561, 1997

    1997

  7. [7]

    Iterative algorithm for reconstruction of entangled states,

    J. ˇReh´aˇcek, Z. Hradil, and M. Je ˇzek, “Iterative algorithm for reconstruction of entangled states,”Physical Review A, vol. 63, no. 4, p. 040303, 2001

    2001

  8. [8]

    Provable compressed sensing quantum state tomography via non-convex methods,

    A. Kyrillidis, A. Kalev, D. Park, S. Bhojanapalli, C. Caramanis, and S. Sanghavi, “Provable compressed sensing quantum state tomography via non-convex methods,”npj Quantum Information, vol. 4, no. 1, p. 36, 2018

    2018

  9. [9]

    Fast and robust quantum state tomography from few basis measurements,

    D. S. Fran c ¸a, F. G. Brand ˜ao, and R. Kueng, “Fast and robust quantum state tomography from few basis measurements,” in16th Conference on the Theory of Quantum Computation, Communication and Cryptography (TQC 2021). Schloss Dagstuhl–Leibniz-Zentrum f ¨ur Informatik, 2021, pp. 7–1

    2021

  10. [10]

    Quantum state tomography via nonconvex riemannian gradient descent,

    M.-C. Hsu, E.-J. Kuo, W.-H. Yu, J.-F. Cai, and M.-H. Hsieh, “Quantum state tomography via nonconvex riemannian gradient descent,”Physical Review Letters, vol. 132, no. 24, p. 240804, 2024

    2024

  11. [11]

    Neural-network quantum state tomography,

    G. Torlai, G. Mazzola, J. Carrasquilla, M. Troyer, R. Melko, and G. Carleo, “Neural-network quantum state tomography,”Nature Physics, vol. 14, no. 5, pp. 447–450, 2018

    2018

  12. [12]

    Machine learning and the physical sciences,

    G. Carleo, I. Cirac, K. Cranmer, L. Daudet, M. Schuld, N. Tishby, L. V ogt-Maranto, and L. Zdeborov ´a, “Machine learning and the physical sciences,” Reviews of Modern Physics, vol. 91, no. 4, p. 045002, 2019

    2019

  13. [13]

    Low rank matrix recovery from rank one measurements,

    R. Kueng, H. Rauhut, and U. Terstiege, “Low rank matrix recovery from rank one measurements,”Applied and Computational Harmonic Analysis, vol. 42, no. 1, pp. 88–116, 2017

    2017

  14. [14]

    Matrix product states represent ground states faithfully,

    F. Verstraete and J. I. Cirac, “Matrix product states represent ground states faithfully,”Physical Review B—Condensed Matter and Materials Physics, vol. 73, no. 9, p. 094423, 2006

    2006

  15. [15]

    Matrix product state representations,

    D. Perez-Garc´ıa, F. Verstraete, M. M. Wolf, and J. I. Cirac, “Matrix product state representations,”Quantum Information and Computation, vol. 7, no. 5-6, pp. 401–430, 2007

    2007

  16. [16]

    Matrix product operator representations,

    B. Pirvu, V . Murg, J. I. Cirac, and F. Verstraete, “Matrix product operator representations,”New Journal of Physics, vol. 12, no. 2, p. 025012, 2010. 23

    2010

  17. [17]

    Efficient description of many-body systems with matrix product density operators,

    J. Guth Jarkovsk `y, A. Moln ´ar, N. Schuch, and J. I. Cirac, “Efficient description of many-body systems with matrix product density operators,”PRX Quantum, vol. 1, no. 1, p. 010304, 2020

    2020

  18. [18]

    Tensor-train decomposition,

    I. V . Oseledets, “Tensor-train decomposition,”SIAM Journal on Scientific Computing, vol. 33, no. 5, pp. 2295–2317, 2011

    2011

  19. [19]

    On manifolds of tensors of fixed tt-rank,

    S. Holtz, T. Rohwedder, and R. Schneider, “On manifolds of tensors of fixed tt-rank,”Numerische Mathematik, vol. 120, no. 4, pp. 701–731, 2012

    2012

  20. [20]

    Efficient quantum state tomography,

    M. Cramer, M. B. Plenio, S. T. Flammia, R. Somma, D. Gross, S. D. Bartlett, O. Landon-Cardinal, D. Poulin, and Y .-K. Liu, “Efficient quantum state tomography,”Nature Communications, vol. 1, no. 1, p. 149, 2010

    2010

  21. [21]

    A scalable maximum likelihood method for quantum state tomography,

    T. Baumgratz, A. N ¨ußeler, M. Cramer, and M. B. Plenio, “A scalable maximum likelihood method for quantum state tomography,”New Journal of Physics, vol. 15, no. 12, p. 125004, 2013

    2013

  22. [22]

    Efficient tomography of a quantum many-body system,

    B. P. Lanyon, C. Maier, M. Holz ¨apfel, T. Baumgratz, C. Hempel, P. Jurcevic, I. Dhand, A. Buyskikh, A. J. Daley, M. Crameret al., “Efficient tomography of a quantum many-body system,”Nature Physics, vol. 13, no. 12, pp. 1158–1162, 2017

    2017

  23. [23]

    Quantum state tomography with locally purified density operators and local measurements,

    Y . Guo and S. Yang, “Quantum state tomography with locally purified density operators and local measurements,”Communications Physics, vol. 7, no. 1, p. 322, 2024

    2024

  24. [24]

    Sketch tomography: Hybridizing classical shadow and matrix product state,

    X. Tang, H. Chen, Y . Khoo, and L. Ying, “Sketch tomography: Hybridizing classical shadow and matrix product state,”arXiv preprint arXiv:2512.03333, 2025

    work page arXiv 2025

  25. [25]

    Density matrix reconstruction using non-negative matrix product states,

    D. Han, C. Guo, and X. Wang, “Density matrix reconstruction using non-negative matrix product states,”Physical Review A, vol. 106, no. 4, p. 042435, 2022

    2022

  26. [26]

    Quantum process tomography with unsupervised learning and tensor networks,

    G. Torlai, C. J. Wood, A. Acharya, G. Carleo, J. Carrasquilla, and L. Aolita, “Quantum process tomography with unsupervised learning and tensor networks,”Nature Communications, vol. 14, no. 1, p. 2858, 2023

    2023

  27. [27]

    Learning the tensor network model of a quantum state using a few single-qubit measurements,

    S. Kuzmin, V . Mikhailova, I. Dyakonov, and S. Straupe, “Learning the tensor network model of a quantum state using a few single-qubit measurements,” Physical Review A, vol. 109, no. 5, p. 052616, 2024

    2024

  28. [28]

    Quantum state tomography for matrix product density operators,

    Z. Qin, C. Jameson, Z. Gong, M. B. Wakin, and Z. Zhu, “Quantum state tomography for matrix product density operators,”IEEE Transactions on Information Theory, vol. 70, no. 7, pp. 5030–5056, 2024

    2024

  29. [29]

    Quantum state tomography via compressed sensing,

    D. Gross, Y .-K. Liu, S. T. Flammia, S. Becker, and J. Eisert, “Quantum state tomography via compressed sensing,”Physical Review Letters, vol. 105, no. 15, p. 150401, 2010

    2010

  30. [30]

    Experimental quantum compressed sensing for a seven-qubit system,

    C. A. Riofrio, D. Gross, S. T. Flammia, T. Monz, D. Nigg, R. Blatt, and J. Eisert, “Experimental quantum compressed sensing for a seven-qubit system,” Nature Communications, vol. 8, no. 1, p. 15305, 2017

    2017

  31. [31]

    An online optimization algorithm for the real-time quantum state tomography: K. zhang et al

    K. Zhang, S. Cong, K. Li, and T. Wang, “An online optimization algorithm for the real-time quantum state tomography: K. zhang et al.”Quantum Information Processing, vol. 19, no. 10, p. 361, 2020

    2020

  32. [32]

    Adaptive online learning of quantum states,

    X. Chen, E. Hazan, T. Li, Z. Lu, X. Wang, and R. Yang, “Adaptive online learning of quantum states,”Quantum, vol. 8, p. 1471, 2024

    2024

  33. [33]

    Online quantum state tomography via stochastic gradient descent,

    J.-F. Cai, Y . Jiao, Y . Li, X. Lu, J. Z. Yang, and J. You, “Online quantum state tomography via stochastic gradient descent,”arXiv preprint arXiv:2507.07601, 2025

    work page arXiv 2025

  34. [34]

    Fast quantum state reconstruction via accelerated non-convex programming,

    J. L. Kim, G. Kollias, A. Kalev, K. X. Wei, and A. Kyrillidis, “Fast quantum state reconstruction via accelerated non-convex programming,” inPhotonics, vol. 10, no. 2. MDPI, 2023, p. 116

    2023

  35. [35]

    Provable tensor-train format tensor completion by riemannian optimization,

    J.-F. Cai, J. Li, and D. Xia, “Provable tensor-train format tensor completion by riemannian optimization,”Journal of Machine Learning Research, vol. 23, no. 123, pp. 1–77, 2022

    2022

  36. [36]

    Fast and provable tensor-train format tensor completion via precondtioned riemannian gradient descent,

    F. Bian, J.-F. Cai, X. Zhang, and Y . Zhang, “Fast and provable tensor-train format tensor completion via precondtioned riemannian gradient descent,” arXiv preprint arXiv:2501.13385, 2025

    work page arXiv 2025

  37. [37]

    M. A. Nielsen and I. L. Chuang,Quantum computation and quantum information. Cambridge university press, 2010

    2010

  38. [38]

    The bloch vector for n-level systems,

    G. Kimura, “The bloch vector for n-level systems,”Physics Letters A, vol. 314, no. 5-6, pp. 339–349, 2003

    2003

  39. [39]

    Optimal entanglement witnesses for qubits and qutrits,

    R. A. Bertlmann, K. Durstberger, B. C. Hiesmayr, and P. Krammer, “Optimal entanglement witnesses for qubits and qutrits,”Physical Review A—Atomic, Molecular, and Optical Physics, vol. 72, no. 5, p. 052331, 2005

    2005

  40. [40]

    Bloch vectors for qudits,

    R. A. Bertlmann and P. Krammer, “Bloch vectors for qudits,”Journal of Physics A: Mathematical and Theoretical, vol. 41, no. 23, p. 235303, 2008

    2008

  41. [41]

    Fast state tomography with optimal error bounds,

    M. Gut ¸˘a, J. Kahn, R. Kueng, and J. A. Tropp, “Fast state tomography with optimal error bounds,”Journal of Physics A: Mathematical and Theoretical, vol. 53, no. 20, p. 204001, 2020

    2020

  42. [42]

    Matrix product density operators: Simulation of finite-temperature and dissipative systems,

    F. Verstraete, J. J. Garcia-Ripoll, and J. I. Cirac, “Matrix product density operators: Simulation of finite-temperature and dissipative systems,”Physical Review Letters, vol. 93, no. 20, p. 207204, 2004

    2004

  43. [43]

    Universal low-rank matrix recovery from pauli measurements,

    Y .-K. Liu, “Universal low-rank matrix recovery from pauli measurements,”Advances in Neural Information Processing Systems, vol. 24, 2011

    2011

  44. [44]

    Sample-optimal tomography of quantum states,

    J. Haah, A. W. Harrow, Z. Ji, X. Wu, and N. Yu, “Sample-optimal tomography of quantum states,” inProceedings of the forty-eighth annual ACM symposium on Theory of Computing, 2016, pp. 913–925

    2016

  45. [45]

    Sample-efficient quantum state tomography for structured quantum states in one dimension,

    Z. Qin, C. Jameson, A. Goldar, M. B. Wakin, Z. Gong, and Z. Zhu, “Sample-efficient quantum state tomography for structured quantum states in one dimension,”arXiv preprint arXiv:2410.02583, 2024

    work page arXiv 2024

  46. [46]

    On tensor completion via nuclear norm minimization,

    M. Yuan and C.-H. Zhang, “On tensor completion via nuclear norm minimization,”Foundations of Computational Mathematics, vol. 16, no. 4, pp. 1031–1068, 2016

    2016

  47. [47]

    Generalized low-rank plus sparse tensor estimation by fast riemannian optimization,

    J.-F. Cai, J. Li, and D. Xia, “Generalized low-rank plus sparse tensor estimation by fast riemannian optimization,”Journal of the American Statistical Association, vol. 118, no. 544, pp. 2588–2604, 2023

    2023

  48. [48]

    Provable tensor factorization with missing data,

    P. Jain and S. Oh, “Provable tensor factorization with missing data,”Advances in Neural Information Processing Systems, vol. 27, 2014

    2014

  49. [49]

    Nonconvex low-rank tensor completion from noisy data,

    C. Cai, G. Li, H. V . Poor, and Y . Chen, “Nonconvex low-rank tensor completion from noisy data,”Advances in Neural Information Processing Systems, vol. 32, 2019

    2019

  50. [50]

    Statistically optimal and computationally efficient low rank tensor completion from noisy entries,

    D. Xia, M. Yuan, and C.-H. Zhang, “Statistically optimal and computationally efficient low rank tensor completion from noisy entries,”The Annals of Statistics, vol. 49, no. 1, pp. 76–99, 2021

    2021

  51. [51]

    Preconditioned riemannian gradient descent algorithm for low-multilinear-rank tensor completion,

    Y . Zhang, F. Bian, X. Zhang, and J.-F. Cai, “Preconditioned riemannian gradient descent algorithm for low-multilinear-rank tensor completion,” in Forty-second International Conference on Machine Learning, 2025

    2025

  52. [52]

    Nonconvex low-rank tensor completion from noisy data,

    C. Cai, G. Li, H. V . Poor, and Y . Chen, “Nonconvex low-rank tensor completion from noisy data,”Operations Research, 2021

    2021

  53. [53]

    Scaling and scalability: Provable nonconvex low-rank tensor estimation from incomplete measurements,

    T. Tong, C. Ma, A. Prater-Bennette, E. Tripp, and Y . Chi, “Scaling and scalability: Provable nonconvex low-rank tensor estimation from incomplete measurements,”Journal of Machine Learning Research, vol. 23, no. 163, pp. 1–77, 2022

    2022

  54. [54]

    Online tensor learning: Computational and statistical trade-offs, adaptivity and optimal regret,

    J. Li, J.-F. Cai, Y . Chen, and D. Xia, “Online tensor learning: Computational and statistical trade-offs, adaptivity and optimal regret,”arXiv preprint arXiv:2306.03372, 2023

    work page arXiv 2023

  55. [55]

    A single-mode quasi riemannian gradient descent algorithm for low-multilinear-rank tensor recovery,

    Y . Zhang, Y .-N. Zhu, and X. Zhang, “A single-mode quasi riemannian gradient descent algorithm for low-multilinear-rank tensor recovery,”Journal of Scientific Computing, vol. 106, no. 1, p. 23, 2026

    2026

  56. [56]

    On polynomial time methods for exact low-rank tensor completion,

    D. Xia and M. Yuan, “On polynomial time methods for exact low-rank tensor completion,”Foundations of Computational Mathematics, vol. 19, no. 6, pp. 1265–1313, 2019

    2019

  57. [57]

    The itensor software library for tensor network calculations,

    M. Fishman, S. White, and E. M. Stoudenmire, “The itensor software library for tensor network calculations,”SciPost Physics Codebases, p. 004, 2022

    2022

  58. [58]

    Matrix completion from a few entries,

    R. H. Keshavan, A. Montanari, and S. Oh, “Matrix completion from a few entries,”IEEE Transactions on Information Theory, vol. 56, no. 6, pp. 2980–2998, 2010

    2010