pith. sign in

arxiv: 2604.14428 · v1 · submitted 2026-04-15 · 🪐 quant-ph · eess.SP

Distributed Learning of Quantum State Tomography Robust to Readout Errors

Amirhossein Taherpour , Alireza Sadeghi , Georgios B. Giannakis This is my paper

Pith reviewed 2026-05-10 12:36 UTC · model grok-4.3

classification 🪐 quant-ph eess.SP
keywords quantum state tomographyreadout errorsoverlapping tomographydistributed optimizationbilinear optimizationADMMscalable quantum estimation
0
0 comments X

The pith

A distributed alternating optimization jointly estimates quantum states and unknown readout errors by enforcing consistency across overlapping subsystems.

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

The paper seeks to make tomography of large multiqubit systems practical when readout devices introduce unknown errors. It partitions the qubits into overlapping regions and assigns each region both a local density operator and a local confusion matrix, then couples neighboring regions by requiring that their reduced states agree on shared subsystems. This coupling produces a bilinear optimization problem that the authors solve with a distributed alternating procedure: ADMM handles the state updates while confusion-matrix updates run locally in parallel. If the method succeeds, it recovers accurate states without needing to calibrate measurements in advance or assume they are error-free, and it does so with provable local identifiability and convergence properties.

Core claim

By coupling local density operators and confusion matrices through reduced-state consistency constraints on overlapping subsystems, the approach yields a structured bilinear optimization problem that admits a distributed alternating solver with analytical guarantees including a sufficient condition for local identifiability, local quadratic growth of the population misfit, and convergence of the inner state-update procedure.

What carries the argument

Reduced-state consistency constraints that couple neighboring regional density operators and confusion matrices into a jointly solvable bilinear optimization problem.

Load-bearing premise

Reduced-state consistency constraints on overlapping subsystems are sufficient to couple the local density operators and confusion matrices into a jointly identifiable bilinear problem without extra regularization or assumptions on the noise model.

What would settle it

A controlled simulation on one of the tested graph geometries in which the joint estimator shows no accuracy gain over separate state estimation with fixed readout matrices, or in which the local identifiability condition fails for the chosen partition.

Figures

Figures reproduced from arXiv: 2604.14428 by Alireza Sadeghi, Amirhossein Taherpour, Georgios B. Giannakis.

Figure 1
Figure 1. Figure 1: A global quantum system partitioned into R local regions. Reduced-state consensus among overlapping regions is reached by partial trace over regional states, offering distributed joint regional state and readout estimation. A. Global-to-Local Decomposition with Overlap Consen￾sus Let ρ ∈ D(Hq) denote the global q-qubit state intro￾duced in Sec. II. To obtain a structured alternative to the global problem (… view at source ↗
Figure 2
Figure 2. Figure 2: Regional graph geometries used in the simulations. Each sub-figure shows a representative overlapping pair of regions. Red nodes denote the shared sites, while green and blue nodes belong only to the left and right region, respectively. In the Hub geometry, the shared sites form the common core. and the normalized state optimality gap g k,ℓ opt := J k ρ [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Effect of the readout regularization parameter λ on the outer alternating updates. terms. Its main role is to judiciously control the updates of the confusion matrices over iterations. When λ is too small, the state error drops quickly at first, but the readout updates become less stable later. When λ is too large, the confusion matrices stay too close to Ir, so both errors decrease more slowly. The best b… view at source ↗
Figure 6
Figure 6. Figure 6: demonstrates the tradeoff between state recovery and algorithmic cost across the four considered graph geometries. The vertical axis is the state error eρ, and the horizontal one is the communication cost defined as Cbud := L¯ X (r,r′)∈O 4 qrr′ , 103 104 Communication budget, Cbud 0.08 0.10 0.12 0.14 0.16 e ρ Ring Ladder Torus Hub (a) Communication tradeoff 107 2 × 107 Computation budget, Wbud 0.08 0.10 0.… view at source ↗
read the original abstract

Scalable estimation of quantum states with readout errors is a central challenge in large multiqubit systems. Existing overlapping-tomography methods improve scalability by working with local subsystems, but they usually assume known or separately calibrated measurements. At the same time, readout-estimation methods model measurement errors without enforcing consistency among overlapping regional states. In this context, the present paper introduces a unified framework for joint regional quantum state tomography with readout errors. A multiqubit system is partitioned in overlapping regions, each region is assigned to a local density operator and a local confusion matrix, and neighboring regions are coupled through reduced-state consistency on shared subsystems. This leads to a structured bilinear optimization problem. To solve it, a distributed alternating method is developed in which the state-update step is handled by the alternating direction method of multipliers (ADMM), while the confusion-matrix updates are carried out locally in parallel. Analytical guarantees are also established, including a sufficient condition for local identifiability, local quadratic growth of the population misfit, and convergence of the inner state-update procedure. Simulations on Ring, Ladder, Torus, and Hub graph geometries show that joint estimation improves state recovery over fixed-readout reconstruction, recovers a substantial portion of oracle performance, and reveals a clear tradeoff between state estimation performance, communication, and computation.

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

3 major / 2 minor

Summary. The paper proposes a unified framework for joint quantum state tomography and readout error estimation in large multiqubit systems. It partitions the system into overlapping regions, assigns each a local density operator ρ_r and confusion matrix C_r, and couples them via reduced-state consistency constraints Tr_{r∖s}(ρ_r) = Tr_{s∖r}(ρ_s). This yields a structured bilinear optimization problem solved by a distributed alternating algorithm (ADMM for state updates, parallel local updates for confusion matrices), with claimed analytical guarantees on local identifiability, quadratic growth of the population misfit, and inner-loop convergence. Simulations on Ring, Ladder, Torus, and Hub graphs demonstrate improved state recovery over fixed-readout methods and partial recovery of oracle performance.

Significance. If the identifiability and convergence results hold under the stated conditions, the work provides a scalable, consistent approach to tomography that integrates readout calibration without separate experiments, potentially improving accuracy in NISQ devices. The distributed ADMM structure and graph-topology simulations are practical strengths; the bilinear formulation with consistency constraints is a novel unification of overlapping tomography and readout modeling.

major comments (3)
  1. [§4] §4 (Analytical Guarantees): The sufficient condition for local identifiability is presented as relying on the reduced-state consistency constraints alone. However, the derivation does not explicitly demonstrate that these marginal equalities eliminate all gauge freedoms (e.g., global scaling or unitary ambiguities) in the bilinear objective between {ρ_r} and {C_r} for arbitrary noise models. A concrete counter-example or additional structural assumption on C_r (such as diagonal dominance) would be needed to confirm the condition is load-bearing and non-vacuous.
  2. [§5] §5 (Simulations): The reported improvements in state recovery (e.g., fidelity gains over fixed-readout baselines) lack error bars, multiple random seeds, or statistical tests across the four graph geometries. Without these, it is difficult to assess whether the observed tradeoff between estimation performance, communication, and computation is robust or sensitive to initialization and hyper-parameters.
  3. [§3] §3 (Problem Formulation): The bilinear objective is non-convex; the local quadratic growth claim for the population misfit is stated but its proof appears to assume the consistency constraints are sufficient to make the feasible set compact and the Hessian positive definite at the true point. An explicit verification that the constraints remove the kernel of the bilinear map would strengthen this.
minor comments (2)
  1. [§2] The notation for reduced-state consistency (Tr_{r∖s}) could be illustrated with a small two-qubit example in the methods section to improve readability.
  2. [Introduction] The abstract claims 'analytical guarantees including a sufficient condition for local identifiability'; the introduction should briefly contrast this with prior overlapping-tomography works that assume known measurements.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive feedback on our manuscript. We address each major comment below and indicate the revisions we plan to make.

read point-by-point responses

  1. Referee: [§4] §4 (Analytical Guarantees): The sufficient condition for local identifiability is presented as relying on the reduced-state consistency constraints alone. However, the derivation does not explicitly demonstrate that these marginal equalities eliminate all gauge freedoms (e.g., global scaling or unitary ambiguities) in the bilinear objective between {ρ_r} and {C_r} for arbitrary noise models. A concrete counter-example or additional structural assumption on C_r (such as diagonal dominance) would be needed to confirm the condition is load-bearing and non-vacuous.

    Authors: We appreciate this observation. In the revised manuscript, we will expand the proof of Theorem 1 to explicitly show how the overlapping consistency constraints eliminate the gauge freedoms. Specifically, we will demonstrate that the reduced-state equalities, combined with the trace-1 normalization on each ρ_r, fix the global scaling, and that the positivity and the structure of the confusion matrices (which are stochastic) prevent unitary ambiguities under the assumed noise model. For arbitrary noise models without additional assumptions, the condition may indeed be vacuous, but our framework assumes standard readout error models where C_r are column-stochastic with positive diagonals. We will add a remark clarifying this and provide a brief counter-example for the case without positivity assumptions. revision: yes

  2. Referee: [§5] §5 (Simulations): The reported improvements in state recovery (e.g., fidelity gains over fixed-readout baselines) lack error bars, multiple random seeds, or statistical tests across the four graph geometries. Without these, it is difficult to assess whether the observed tradeoff between estimation performance, communication, and computation is robust or sensitive to initialization and hyper-parameters.

    Authors: We agree that the simulation results would benefit from statistical validation. In the revised version, we will rerun the experiments with 10 random seeds for each graph geometry, include error bars representing standard deviations, and add a brief statistical analysis (e.g., paired t-tests) to confirm the significance of the improvements. We will also discuss sensitivity to initialization and hyperparameters in a new subsection. revision: yes

  3. Referee: [§3] §3 (Problem Formulation): The bilinear objective is non-convex; the local quadratic growth claim for the population misfit is stated but its proof appears to assume the consistency constraints are sufficient to make the feasible set compact and the Hessian positive definite at the true point. An explicit verification that the constraints remove the kernel of the bilinear map would strengthen this.

    Authors: Thank you for pointing this out. The proof of the quadratic growth in Proposition 2 relies on the local identifiability from Theorem 1, which ensures that the constraints remove the kernel at the true point. However, we acknowledge that the compactness of the feasible set and positive-definiteness of the Hessian are not fully detailed. In the revision, we will add an explicit verification step showing that the consistency constraints, together with the boundedness of the density operators and stochastic matrices, make the feasible set compact, and that the second derivative (Hessian) is positive definite in the tangent space orthogonal to the constraints at the true solution. This will be included as a lemma supporting the quadratic growth claim. revision: yes

Circularity Check

0 steps flagged

No significant circularity; analytical guarantees are independently derived

full rationale

The paper constructs a bilinear optimization problem directly from the reduced-state consistency constraints on overlapping subsystems, then derives a distributed ADMM solver and states sufficient conditions for local identifiability, quadratic growth, and convergence. These guarantees are presented as established analytical results rather than tautological restatements or fits to the same inputs. No load-bearing self-citations, self-definitional loops, or renamed known results appear in the derivation chain. The framework adds independent content via the joint estimation procedure and its convergence analysis, making the overall chain self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no explicit free parameters, axioms, or invented entities; the framework itself is the main contribution but its internal assumptions cannot be audited without the full text.

pith-pipeline@v0.9.0 · 5536 in / 1001 out tokens · 25658 ms · 2026-05-10T12:36:38.450708+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

  1. [1]

    M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, 10th ed. Cambridge: Cambridge Uni- versity Press, 2010

    work page 2010

  2. [2]

    Watrous, The Theory of Quantum Information

    J. Watrous, The Theory of Quantum Information. Cambridge: Cambridge University Press, 2018

    work page 2018

  3. [3]

    M. G. A. Paris and J. Řeháček, Eds., Quantum State Estima- tion, ser. Lecture Notes in Physics. Berlin, Heidelberg: Springer, 2004, vol. 649

    work page 2004

  4. [4]

    Symmetric informationally complete quantum measurements,

    J. M. Renes, R. Blume-Kohout, A. J. Scott, and C. M. Caves, “Symmetric informationally complete quantum measurements,” Journal of Mathematical Physics, vol. 45, no. 6, pp. 2171–2180, 2004

    work page 2004

  5. [5]

    Quantum overlapping tomography,

    J. Cotler and F. Wilczek, “Quantum overlapping tomography,” Physical Review Letters, vol. 124, no. 10, p. 100401, 2020

    work page 2020

  6. [6]

    Predicting many properties of a quantum system from very few measurements,

    H.-Y. Huang, R. Kueng, and J. Preskill, “Predicting many properties of a quantum system from very few measurements,” Nature Physics, vol. 16, no. 10, pp. 1050–1057, 2020

    work page 2020

  7. [7]

    Local quantum overlapping tomography,

    B. G. M. Araújo, M. M. Taddei, D. Cavalcanti, and A. Acín, “Local quantum overlapping tomography,” Physical Review A, vol. 106, no. 6, p. 062441, 2022

    work page 2022

  8. [8]

    Experimental sample-efficient quantum state tomography via parallel measurements,

    C.-K. Hu, C. Wei, C. Liu, L. Che, Y. Zhou, G. Xie, H. Qin, G. Hu, H. Yuan, R. Zhou, S. Liu, D. Tan, T. Xin, and D. Yu, “Experimental sample-efficient quantum state tomography via parallel measurements,” Physical Review Letters, vol. 133, no. 16, p. 160801, 2024

    work page 2024

  9. [9]

    Optimal overlapping tomography,

    K. Hansenne, R. Qu, L. T. Weinbrenner, C. de Gois, H. Wang, Y. Ming, Z. Yang, P. Horodecki, W. Gao, and O. Gühne, “Optimal overlapping tomography,” Physical Review Letters, vol. 135, no. 6, p. 060801, 2025

    work page 2025

  10. [10]

    Two-stage estimation for quantum de- tector tomography: Error analysis, numerical and experimental results,

    Y. Wang, S. Yokoyama, D. Dong, I. R. Petersen, E. H. Hunting- ton, and H. Yonezawa, “Two-stage estimation for quantum de- tector tomography: Error analysis, numerical and experimental results,” IEEE Trans. Inf. Theory, vol. 67, no. 4, pp. 2293–2307, 2021

    work page 2021

  11. [11]

    Unrolling SVT to obtain computationally efficient SVT for N-qubit quantum state to- mography,

    S. Shanmugam and S. Kalyani, “Unrolling SVT to obtain computationally efficient SVT for N-qubit quantum state to- mography,” IEEE Trans. Signal Process., vol. 71, pp. 178–188, 2023

    work page 2023

  12. [12]

    Generalized quantum state tomography with hybrid denoising priors,

    D. Xue, W. Dai, Z. Zheng, C. Li, J. Zou, and H. Xiong, “Generalized quantum state tomography with hybrid denoising priors,” IEEE Trans. Signal Process., vol. 73, pp. 1532–1548, 2025

    work page 2025

  13. [13]

    Low-rank positive semidefinite matrix recovery from corrupted rank-one measurements,

    Y. Li, Y. Sun, and Y. Chi, “Low-rank positive semidefinite matrix recovery from corrupted rank-one measurements,” IEEE Trans. Signal Process., vol. 65, no. 2, pp. 397–408, 2017

    work page 2017

  14. [14]

    Low-rank matrix recovery with scaled subgradient methods: Fast and robust convergence without the condition number,

    T. Tong, C. Ma, and Y. Chi, “Low-rank matrix recovery with scaled subgradient methods: Fast and robust convergence without the condition number,” IEEE Trans. Signal Process., vol. 69, pp. 2396–2409, 2021

    work page 2021

  15. [15]

    Plug- and-play priors for model based reconstruction,

    S. V. Venkatakrishnan, C. A. Bouman, and B. Wohlberg, “Plug- and-play priors for model based reconstruction,” in Proc. IEEE Global Conf. Signal Inf. Process., 2013, pp. 945–948

    work page 2013

  16. [16]

    Regularization by denoising: Clarifications and new interpretations,

    E. T. Reehorst and P. Schniter, “Regularization by denoising: Clarifications and new interpretations,” IEEE Trans. Comput. Imag., vol. 5, no. 1, pp. 52–67, 2019

    work page 2019

  17. [17]

    On the convergence of non-convex phase retrieval with denoising priors,

    D. Xue, Z. Zheng, W. Dai, C. Li, J. Zou, and H. Xiong, “On the convergence of non-convex phase retrieval with denoising priors,” IEEE Trans. Signal Process., vol. 70, pp. 4424–4439, 2022

    work page 2022

  18. [18]

    An overview of advances in signal processing techniques for classical and quantum wideband synthetic apertures,

    P. Vouras, K. V. Mishra, A. B. Artusio-Glimpse, S. Pinilla, A. Xenaki, D. W. Griffith, and K. Egiazarian, “An overview of advances in signal processing techniques for classical and quantum wideband synthetic apertures,” IEEE J. Sel. Topics Signal Process., vol. 17, no. 2, pp. 317–369, 2023

    work page 2023

  19. [19]

    Widely distributed radar imaging: Unmediated ADMM based approach,

    A. Murtada, R. Hu, B. S. Mysore Rama Rao, and U. Schroeder, “Widely distributed radar imaging: Unmediated ADMM based approach,” IEEE J. Sel. Topics Signal Process., vol. 17, no. 2, pp. 389–402, 2023

    work page 2023

  20. [20]

    Distributed robust power system state estimation,

    V. Kekatos and G. B. Giannakis, “Distributed robust power system state estimation,” IEEE Trans. Power Syst., vol. 28, no. 2, pp. 1617–1626, 2013

    work page 2013

  21. [21]

    Mitigation of readout noise in near-term quantum devices by classical post- processing based on detector tomography,

    F. B. Maciejewski, Z. Zimboras, and M. Oszmaniec, “Mitigation of readout noise in near-term quantum devices by classical post- processing based on detector tomography,” Quantum, vol. 4, p. 257, 2020

    work page 2020

  22. [22]

    Unfolding quantum computer readout noise,

    B. Nachman, M. Urbanek, W. A. de Jong, and C. W. Bauer, “Unfolding quantum computer readout noise,” npj Quantum Information, vol. 6, p. 84, 2020

    work page 2020

  23. [23]

    Mitigating measurement errors in multiqubit exper- iments,

    S. Bravyi, S. Sheldon, A. Kandala, D. C. McKay, and J. M. Gambetta, “Mitigating measurement errors in multiqubit exper- iments,” Physical Review A, vol. 103, no. 4, p. 042605, 2021

    work page 2021

  24. [24]

    Scalable mitigation of measurement errors on quantum com- puters,

    P. D. Nation, H. Kang, N. Sundaresan, and J. M. Gambetta, “Scalable mitigation of measurement errors on quantum com- puters,” PRX Quantum, vol. 2, no. 4, p. 040326, 2021

    work page 2021

  25. [25]

    Readout error mitigated quantum state to- mography tested on superconducting qubits,

    A. S. Aasen, A. D. Giovanni, H. Rotzinger, A. V. Ustinov, and M. Gärttner, “Readout error mitigated quantum state to- mography tested on superconducting qubits,” Communications Physics, vol. 7, p. 301, 2024

    work page 2024

  26. [26]

    Mitigation of correlated readout errors without random- ized measurements,

    ——, “Mitigation of correlated readout errors without random- ized measurements,” Physical Review A, vol. 112, no. 2, p. 022601, 2025

    work page 2025

  27. [27]

    Gate set tomography,

    E. Nielsen, J. K. Gamble, K. Rudinger, T. Scholten, K. Young, and R. Blume-Kohout, “Gate set tomography,” Quantum, vol. 5, p. 557, 2021

    work page 2021

  28. [28]

    Self- consistent quantum measurement tomography based on semidef- 15 inite programming,

    M. Cattaneo, M. A. C. Rossi, K. Korhonen, E.-M. Bor- relli, G. García-Pérez, Z. Zimborás, and D. Cavalcanti, “Self- consistent quantum measurement tomography based on semidef- 15 inite programming,” Physical Review Research, vol. 5, no. 3, p. 033154, 2023

    work page 2023

  29. [29]

    Universal framework for simultaneous tomogra- phy of quantum states and SPAM noise,

    A. Jayakumar, S. Chessa, C. Coffrin, A. Y. Lokhov, M. Vuffray, and S. Misra, “Universal framework for simultaneous tomogra- phy of quantum states and SPAM noise,” Quantum, vol. 8, p. 1426, 2024

    work page 2024

  30. [30]

    Inequalities for the L1 deviation of the empirical distribution,

    T. Weissman, E. Ordentlich, G. Seroussi, S. Verdu, and M. J. Weinberger, “Inequalities for the L1 deviation of the empirical distribution,” HP Laboratories, Tech. Rep. HPL-2003-97(R.1), 2003

    work page 2003

  31. [31]

    Exper- imental demonstration of quantum overlapping tomography,

    Z. Yang, S. Ru, L. Cao, N. Zheludev, and W. Gao, “Exper- imental demonstration of quantum overlapping tomography,” Physical Review Letters, vol. 130, no. 5, p. 050804, 2023

    work page 2023

  32. [32]

    Optimal quantum overlapping tomography: Theory and experiment,

    C. Wei, K. Yang, L. Che, F. Xu, J. Song, and T. Xin, “Optimal quantum overlapping tomography: Theory and experiment,” Physical Review Applied, vol. 24, no. 4, p. 044091, 2025

    work page 2025

  33. [33]

    Mitigating shot noise in local overlapping quantum tomography with semidefinite programming,

    Z. J. Wang, D. Dechant, Y. J. Patel, and J. Tura, “Mitigating shot noise in local overlapping quantum tomography with semidefinite programming,” Physical Review A, vol. 111, no. 5, p. 052444, 2025

    work page 2025

  34. [34]

    Steeb and Y

    W.-H. Steeb and Y. Hardy, Problems and Solutions in Quantum Computing and Quantum Information, 2nd ed. Singapore: World Scientific, 2006

    work page 2006

  35. [35]

    Qdataset, quantum datasets for machine learning,

    E. Perrier, A. Youssry, and C. Ferrie, “Qdataset, quantum datasets for machine learning,” Scientific Data, vol. 9, p. 582, 2022

    work page 2022

  36. [36]

    Qdataset: Quantum datasets for machine learning,

    ——, “Qdataset: Quantum datasets for machine learning,” 2021, gitHub repository, version 1.0.0. [Online]. A vailable: https://github.com/eperrier/QDataSet

    work page 2021

  37. [37]

    Qiskit experi- ments: A python package to characterize and calibrate quantum computers,

    N. Kanazawa, D. J. Egger, Y. Ben-Haim, H. Zhang, W. E. Shanks, G. Aleksandrowicz, and C. J. Wood, “Qiskit experi- ments: A python package to characterize and calibrate quantum computers,” Journal of Open Source Software, vol. 8, no. 84, p. 5329, 2023

    work page 2023

  38. [38]

    Quantum state tomography,

    Qiskit Experiments developers, “Quantum state tomography,” Qiskit Experiments documentation, version 0.13.0, 2025, accessed for standard tomographically complete measurement workflow and built-in tomography basis classes. [Online]. A vail- able: https://qiskit-community.github.io/qiskit-experiments/ manuals/verification/state_tomography.html

    work page 2025

  39. [39]

    Probing quantum processor performance with pygsti,

    E. Nielsen, K. Rudinger, T. Proctor, A. Russo, K. Young, and R. Blume-Kohout, “Probing quantum processor performance with pygsti,” Quantum Science and Technology, vol. 5, no. 4, p. 044002, 2020

    work page 2020

  40. [40]

    Dis- tributed optimization and statistical learning via the alternating direction method of multipliers,

    S. Boyd, N. Parikh, E. Chu, B. Peleato, and J. Eckstein, “Dis- tributed optimization and statistical learning via the alternating direction method of multipliers,” Foundations and Trends in Machine Learning, vol. 3, no. 1, pp. 1–122, 2011

    work page 2011

  41. [41]

    On the Douglas–Rachford splitting method and the proximal point algorithm for maximal monotone operators,

    J. Eckstein and D. P. Bertsekas, “On the Douglas–Rachford splitting method and the proximal point algorithm for maximal monotone operators,” Mathematical Programming, vol. 55, no. 1–3, pp. 293–318, 1992

    work page 1992