Pith. sign in

REVIEW 3 major objections 5 minor 2 cited by

Learning Circuits with Infinite Tensor Networks

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read One translation-invariant cell of two-qubit gates, optimized in the thermodynamic limit against infinite matrix product states, reproduces time-evolution unitary with fewer CNOT and T gates than standard Trotterization at matched accuracy.

desk verdict Useful new iMPS-based circuit compilation with real Trotter improvements, but the unitary-compression guarantee rests on an unproven extension of an OOD generalization theorem—worth peer review with that as the focus. read the letter →

arxiv 2506.02105 v1 pith:L5SKJD7P submitted 2025-06-02 quant-ph cond-mat.str-el

Joe Gibbs , Lukasz Cincio This is my paper · ORCID
classification quant-phcond-mat.str-el
keywords infinitematrixproductstatesHamiltoniansimulationquantumcircuitcompilationtranslationinvariancetimeevolutionT-countreductiongroundstatepreparationunitarysynthesis
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper claims that the main obstacle to quantum Hamiltonian simulation—gate count—can be attacked by compiling circuits directly in the thermodynamic limit. The authors optimize a single translation-invariant unit cell of two-qubit gates so that the infinite circuit it generates matches a target state computed as an infinite matrix product state (iMPS) in local fidelity; because the unit cell is the only free object, the optimization cost no longer scales with system size. They report that the resulting circuits prepare transverse-field Ising ground states with error that decays exponentially in the number of layers, evolve states under the Thirring and XXZ models with an order of magnitude or more lower error than equal-depth Trotterization, and reproduce the full propagator $e^{-iHt}$ on up to 30 qubits with lower error than equivalent-depth Trotter circuits. After decomposing rotations to the Clifford+T gate set, the compressed circuits reach a target error with up to a factor of 5.2 fewer T gates per qubit. If these results hold at larger scales, the method gives a practical route to shorter near-term circuits and lower fault-tolerant overhead.

What carries the argument

The machinery has four parts. First, the iMPS: a tensor-network ansatz for translation-invariant infinite one-dimensional states whose transfer matrix has a leading eigenvalue of unit magnitude when normalized, and whose leading eigenvalue for a mixed transfer matrix between two states defines the local fidelity used everywhere as the cost. Second, the parameterized circuit: alternating odd/even layers of identical two-qubit $\mathrm{SU}(4)$ gates with 15 parameters each, laid down as a unit cell that can be contracted into an iMPS by iTEBD. Third, the gradient engine: for each layer, the network above and below is contracted into environment iMPS, so the gradient of the cost with respect to that layer can be finite-differenced with warm-started Arnoldi iterations at $O(\chi^2)$ memory. Fourth, the unitary-compilation justification: the out-of-distribution generalization result of Ref. [28], which says a circuit matched on low-entangled training states approximates the full unitary, enabling a handful of random bond-dimension-2 iMPS to serve as the training set.

What would settle it

Run the appendix's generalization protocol on a case the paper does not cover: compile $e^{-iHt}$ for the XXZ Hamiltonian, or for the transverse-field Ising model past $t=3.0$ with 20 or more layers, using four bond-dimension-2 training states, then measure the cost on 100 unseen states. If the training cost keeps dropping while the testing cost plateaus (the divergence shown for a single training state), the out-of-distribution generalization has failed and the compiled unit cell does not realize the unitary. A second check: take the T-count-optimized circuit with 795 T gates per qubit and verify by exact statevector simulation that its infidelity to $e^{-iHt}$ actually reaches the claimed $3\times10^{-7}$ on 30 qubits, and that the error does not grow with system size.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that compiling directly in the infinite limit—rather than on finite systems—is both possible and useful for translation-invariant one-dimensional simulation. The target state or unitary is represented as an iMPS computed by an infinite tensor network algorithm such as iTEBD; the variational circuit is a product of alternating layers of $\mathrm{SU}(4)$ gates defined by a single two-qubit unit cell; and the cost function is the local fidelity given by the leading eigenvalue of the mixed transfer matrix. For unitary compression, the out-of-distribution generalization theorem of Ref. [28] is invoked to justify training on a small set of low-entangled random iMPS, and the paper reports that four training states of bond dimension 2 suffice up to 29 layers. The central quantitative claims are: exponential error decay in the number of layers for ground-state preparation, with the decay rate governed by the correlation length; over an order of magnitude error reduction versus equal-depth Trotterization for time-evolved states; and a 5.2-fold T-count reduction for realizing $e^{-iHt}$ after Clifford+T decomposition, validated by statevector simulation on up to 30 qubits.

Load-bearing premise

The pipeline assumes that a circuit optimized on just four simple, low-entanglement training states will approximate the full time-evolution unitary everywhere in Hilbert space, even though the cited theory guarantees this only for finite systems and calls for far more training states.

Editorial extensions

If this is right

  • A circuit compiled once on an infinite unit cell can be tiled onto any finite system size with no further optimization, so the compilation cost is paid once per Hamiltonian rather than per qubit count.
  • At matched target error, the compiled circuits consume fewer CNOT layers than first-, second-, or fourth-order Trotterizations, which translates directly into smaller two-qubit-gate error budgets on near-term hardware.
  • Constraining the unit cell to a few rotation gates during optimization yields, after Clifford+T decomposition, up to 5.2-fold fewer T gates per qubit than the best equal-error Trotter circuit, reducing magic-state distillation overhead on fault-tolerant machines.
  • Ground-state preparation error decays exponentially with the number of layers at a rate set by the correlation length (approximately $\alpha_1 \propto \xi^{-0.6}$), so even states near the critical point converge systematically as depth grows.
  • The optimized unit cells keep their advantage when transferred to periodic-boundary finite systems up to 30 qubits, a regime where direct finite-MPS compilation is itself costly.

Reading between the lines

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

  • A natural extrapolation the paper does not test is that the same four-state training recipe compiles other translation-invariant short-range propagators, such as XXZ time evolution, Floquet operators, or longer evolution times, where the Trotter baseline is no better and the relative savings could be larger.
  • The 5.2-fold T-count figure is best read as a floor: the authors decompose rotations with the generic Ross-Selinger algorithm after optimization and note that T-count-aware search or better synthesis should improve the ratio.
  • Because the optimization cost is set by the unit cell rather than the system size, the practical ceiling of the approach is the bond dimension and evolution time at which the target iMPS itself stops being classically simulable; past that point the pipeline would need another source of training data.
  • The measured decay of the error rate with correlation length suggests the circuit format loses efficiency exactly where entanglement grows, so a length-scale-adapted ansatz (larger unit cells or growing bond dimension near criticality) is the natural follow-up.
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

3 major / 5 minor

Summary. The paper introduces a method to compile translation-invariant quantum circuits using infinite Matrix Product States (iMPS). The approach uses iMPS datasets generated by iTEBD to define local-fidelity cost functions for ground-state preparation, real-time state evolution, and unitary compression, and optimizes the parameters of the unit cell of an infinite circuit. The authors demonstrate numerically that the optimized circuits outperform Trotterized circuits of equal depth for the TFIM, Thirring, and XXZ models, with a reported 5.2x reduction in T-count for fault-tolerant implementation of e^{-iHt}. A final section shows that circuits optimized in the infinite limit maintain an advantage in finite-size simulations with periodic boundary conditions.

Significance. If the results hold, the paper establishes a practical framework for using infinite tensor networks to compile translation-invariant circuits for Hamiltonian simulation, with substantial reductions in CNOT depth and T-count compared with Trotterization across several models. The numerical evidence is broad, covering ground-state preparation, real-time evolution, unitary compression, Clifford+T decomposition, and finite-size validation, and the method's cost scales with unit-cell size rather than system size, which is a genuine conceptual advance. The authors also provide a useful appendix on generalization as a function of training set size. However, the headline claims for unitary compression and T-count reduction depend on an out-of-distribution generalization assumption that is not fully verified; the manuscript does not ship code, but the algorithmic steps are described in sufficient detail to be reproduced.

major comments (3)
  1. [Section III C, Eq. (12), Fig. 5, Appendix A] The claim that the optimized unit cell faithfully realizes e^{-iHt} rests on the out-of-distribution generalization theorem of Ref. [28], but the manuscript does not establish the theorem's applicability to this setting. The theorem is stated for finite-dimensional systems and global fidelity, whereas the cost function Eq. (12) uses local fidelity on infinite iMPS; the paper gives no proof that the O(L log L) sample-complexity bound transfers. Moreover, the numerical generalization checks (Fig. 5 and Appendix A) test the compiled circuit on NTest=100 random iMPS with chiTest=2, which is the same bond dimension as the chiTrain=2 training distribution. This is an in-distribution test and cannot detect failure on the highly entangled states the theorem was intended to cover. The Haar-random finite validation in Fig. 7 does provide out-of-distribution evidence, but only for the TFIM Hamiltonian and circuit depths 7-11 with n<=30; it does not cover the 29-layer Thirring circuit or the circuits underlying the 5.2x T-count claim. If generalization fails for larger circuits or other Hamiltonians, the central unitary-compression claim is unsupported.
  2. [Section III D, Fig. 6] The T-count reduction of up to 5.2x is evaluated with the approximation error C3 defined in Eq. (12), i.e., the local infidelity on random iMPS with chi=2. This measures performance on the low-entangled training-type distribution, not the global approximation error of the unitary e^{-iHt}. Therefore, the fault-tolerant advantage is conditional on the same out-of-distribution generalization property as Sec. III C. The authors should either provide a valid OOD validation for the specific circuits used in Fig. 6 (including the T-count-optimized TFIM circuit) or explicitly soften the claim that the optimized circuits 'realize the unitary e^{-iHt}' to a claim about low-entangled inputs.
  3. [Section III A, B; Figs. 3 and 4] The quantitative comparisons (exponential decay rates alpha1 and alpha2, and the error-reduction factors of 32.3x and 35.7x) are based on single optimization runs with no reported error bars, multiple seeds, or statistical uncertainty. Without this information, the reader cannot assess whether the observed differences between optimized and Trotterized circuits are significant beyond a single realization. I would ask the authors to report the number of independent runs and the spread of the metrics, or to provide a reproducibility statement.
minor comments (5)
  1. [Section III C] The text says 'we use χTrain = 2 as this larger set of states has improved generalization properties'; here χTrain is a bond dimension, not a number of training states, so the wording should be, for example, 'states with this larger bond dimension'.
  2. [Section I] The 'Note' about Sokolov and Dziarmaga interrupts the introduction; it would be more appropriate as a footnote or in the related-work discussion.
  3. [Section II C, Eq. (2)] The phrase 'minus the identity element' is imprecise; the generator set is the 15 non-identity Pauli operators in {I,X,Y,Z}^{⊗2}.
  4. [General] The manuscript does not state whether code and data are available; for a numerical methods paper, a data/code availability statement would improve reproducibility.
  5. [Fig. 3 caption] The panel labels appear to be duplicated ('a) b)' listed twice); please label the four panels c) and d) explicitly in the caption text.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: all optimization targets and benchmarks are external to the trained circuit, and the cited generalization theorem is independent support rather than a definitional input.

full rationale

The paper's central pipeline optimizes a translation-invariant circuit by minimizing local infidelity cost functions (Eqs. 10-12) against target iMPS states computed by iTEBD. These target states are produced by a separate classical algorithm, not by the circuit being trained, and no parameter is fitted to the final reported error values. The unitary-compression claim relies on the out-of-distribution generalization theorem of Ref. [28], but that theorem is an external mathematical result with its own stated assumptions; despite overlapping authorship, it is not derived from or equivalent to the present paper's cost function. The paper also provides numerical generalization checks on unseen test states, and the finite-system validation in Fig. 7 uses Haar-random states, an independent out-of-distribution benchmark. The T-count comparison in Fig. 6 applies the same Ross-Selinger decomposition and the same cost function to both iMPS-optimized and Trotterized circuits, so the reported 5.2x reduction is not an artifact of using different metrics. The main caveat—that NTrain=4 is below the O(L log L) bound and the extension of Ref. [28] to infinite iMPS is not proven—is a correctness and generalization risk, not circularity, because no equation in the paper reduces to its own input by construction.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The method introduces no new physical entities or fitted parameters in the core algorithm. The free parameters listed are descriptive fits and the hand-chosen training set size. The main assumptions are standard iMPS properties and the validity of the generalization theorem.

free parameters (3)
  • Exponential decay constants alpha1, alpha2 = alpha1 in [0.508, 0.603] depending on metric
    Fitted from exponential fits to the error vs depth data for ground state preparation (Fig. 3). They are descriptive, not used in the central claim.
  • Polynomial exponents for alpha vs correlation length = -0.603 and -0.508
    Fitted power-law relations between the decay constants and correlation length (Fig. 3c,d). Exploratory and not load-bearing.
  • Number of training states NTrain = 4
    Chosen empirically as sufficient for generalization. Not derived from the theory, which suggests O(L log L) states.
assumptions (4)
  • domain assumption The target states and unitaries can be accurately represented as iMPS with bond dimension up to 128 and a 2-site unit cell.
    Used to compute target states via iTEBD. The paper states energy errors are below 1e-8 for TFIM, but this is not proven for all cases.
  • standard math The leading eigenvalue of the mixed transfer matrix provides a faithful local fidelity measure for infinite systems.
    Standard in iMPS literature; the paper cites Vidal's work.
  • domain assumption The out-of-distribution generalization theorem of Ref. [28] applies to the infinite-limit unitary compilation with the chosen training states.
    The paper uses 4 training states, far below the O(L log L) bound. The theorem's assumptions (finite dimensions, distribution of training states) may not directly extend to infinite systems.
  • standard math Accurate finite-system validation via Haar random states approximates the true unitary error.
    Standard result: inner products with random states give an estimate of the Hilbert-Schmidt norm, with the approximation improving exponentially in qubit count.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning Circuits with Infinite Tensor Networks." pith.science (2026). https://pith.science/paper/L5SKJD7P

@misc{pith2026250602105,
  author       = {Pith},
  title        = {Pith review of: Learning Circuits with Infinite Tensor Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L5SKJD7P}},
  note         = {Machine review of arXiv:2506.02105}
}
abstract

Hamiltonian simulation on quantum computers is strongly constrained by gate counts, motivating techniques to reduce circuit depths. While tensor networks are natural competitors to quantum computers, we instead leverage them to support circuit design, with datasets of tensor networks enabling a unitary synthesis inspired by quantum machine learning. For a target simulation in the thermodynamic limit, translation invariance is exploited to significantly reduce the optimization complexity, avoiding a scaling with system size. Our approach finds circuits to efficiently prepare ground states, and perform time evolution on both infinite and finite systems with substantially lower gate depths than conventional Trotterized methods. In addition to reducing CNOT depths, we motivate similar utility for fault-tolerant quantum algorithms, with a demonstrated $5.2\times$ reduction in $T$-count to realize $e^{-iHt}$. The key output of our approach is the optimized unit-cell of a translation invariant circuit. This provides an advantage for Hamiltonian simulation of finite, yet arbitrarily large, systems on real quantum computers.

Figures

Figures reproduced from arXiv: 2506.02105 by the authors.

Figure 1
Figure 1. FIG. 1 [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: shows the result of this calculation. The opti￾mized gates found in the infinite limit are used to cre￾ate 1D translation invariant circuits to realize the time evolution operators for the n-qubit TFIM Hamiltonian with PBC, represented by the data points with colored m…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 2 Pith papers

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

  1. Resource-Efficient Simulations of Particle Scattering on a Digital Quantum Computer

    quant-ph 2025-07 conditional novelty 6.0 of 10

    A hybrid tensor-network plus quantum-hardware pipeline simulates Thirring-model fermion scattering on 40 qubits and prepares wave packets on 80 qubits with a 3.2x circuit depth reduction.

  2. Numerical Experiments with Parameter Setting of Trotterized Quantum Phase Estimation for Quantum Hamiltonian Ground State Computation

    quant-ph 2026-02 conditional novelty 4.0 of 10

    On a 3-qubit Heisenberg spin glass, Trotterized QPE samples the ground-state-energy phase at a rate fixed by initial-state overlap times the textbook QPE success probability, saturating at surprisingly high Trotter error.

Reference graph

Works this paper leans on

42 extracted references · 18 canonical work pages · cited by 2 Pith papers

  1. [28]

    Gover, V

    L. Gover, V. Wimalaweera, F. Azad, M. DeCross, M. Foss-Feig, and A. G. Green, Fully optimised varia- tional simulation of a dynamical quantum phase transi- tion on a trapped-ion quantum computer, arXiv preprint arXiv:2502.06961 10.48550/arXiv.2502.06961 (2025)

    arXiv doi:10.48550/arxiv.2502.06961 2025

  2. [1]

    (7) Throughout our analysis we fix J = 1.0

    The 1D Transverse-Field Ising model (TFIM), de- fined by H = −J X i∈Z σz i σz i+1 − h X i∈Z σx i . (7) Throughout our analysis we fix J = 1.0

    work page

  3. [2]

    The massive Thirring model [32] is a theoretical model of a (1+1)D quantum field theory. After a lattice discretization in the Kogut-Susskind staggered formula- tion [33], it can be represented by the following spin- 1 2 Hamiltonian H = i 2a X i∈Z (σ− i+1σ+ i − σ− i σ+ i+1) + m 2 X i∈Z (−1)i(1 − σz i ) + g 4a X i∈Z (1 − σz i )(1 − σz i+1) . (8) Throughout...

    work page

  4. [3]

    odd” and “even

    The 1D XXZ model, defined by H = X i∈Z Jx(σx i σx i+1 + σy i σy i+1) + Jzσz i σz i+1 . (9) Throughout our analysis we fix Jx = 1.0 and Jz = 0.5. A. Ground-State Preparation Our first demonstration is a circuit compilation to prepare the ground state of the Transverse-Field Ising Hamiltonian, defined in Eq. (7). The system undergoes a phase transition from...

    work page 2000

  5. [4]

    R. P. Feynman, Simulating physics with computers, International Journal of Theoretical Physics 21, 467 (1982)

    1982

  6. [5]

    Preskill, Quantum computing in the NISQ era and beyond, Quantum 2, 79 (2018)

    J. Preskill, Quantum computing in the NISQ era and beyond, Quantum 2, 79 (2018)

    2018

  7. [6]

    Bravyi and J

    S. Bravyi and J. Haah, Magic-state distillation with low overhead, Physical Review A—Atomic, Molecular, and Optical Physics 86, 052329 (2012)

    work page 2012

  8. [7]

    J. Lee, D. W. Berry, C. Gidney, W. J. Huggins, J. R. Mc- Clean, N. Wiebe, and R. Babbush, Even more efficient quantum computations of chemistry through tensor hy- percontraction, PRX Quantum 2, 030305 (2021)

    2021

Show all 42 references
  1. [8]

    Caesura, C

    A. Caesura, C. L. Cortes, W. Pol, S. Sim, M. Steudtner, G.-L. R. Anselmetti, M. Degroote, N. Moll, R. Santa- gati, M. Streif, et al., Faster quantum chemistry simula- tions on a quantum computer with improved tensor fac- torization and active volume compilation, arXiv preprint...

    2025 doi arXiv

  2. [9]

    F¨ osel, M

    T. F¨ osel, M. Y. Niu, F. Marquardt, and L. Li, Quantum circuit optimization with deep reinforce- ment learning, arXiv preprint arXiv:2103.07585 10.48550/arXiv.2103.07585 (2021)

    2021 doi arXiv

  3. [10]

    F. J. Ruiz, T. Laakkonen, J. Bausch, M. Balog, M. Barekatain, F. J. Heras, A. Novikov, N. Fitzpatrick, B. Romera-Paredes, J. van de Wetering, et al., Quantum circuit optimization with alphatensor, Nature Machine Intelligence , 1 (2025)

    2025

  4. [11]

    Duncan, A

    R. Duncan, A. Kissinger, S. Perdrix, and J. Van De We- tering, Graph-theoretic simplification of quantum cir- cuits with the ZX-calculus, Quantum 4, 279 (2020)

    2020

  5. [12]

    Or´ us, Tensor networks for complex quantum systems, Nature Reviews Physics 1, 538 (2019)

    R. Or´ us, Tensor networks for complex quantum systems, Nature Reviews Physics 1, 538 (2019)

    2019

  6. [13]

    Vidal, Efficient classical simulation of slightly entan- gled quantum computations, Phys

    G. Vidal, Efficient classical simulation of slightly entan- gled quantum computations, Phys. Rev. Lett. 91, 147902 (2003)

    2003

  7. [14]

    Berezutskii, A

    A. Berezutskii, A. Acharya, R. Ellerbrock, J. Gray, R. Haghshenas, Z. He, A. Khan, V. Kuzmin, M. Liu, D. Lyakh, et al. , Tensor networks for quantum computing, arXiv preprint arXiv:2503.08626 10.48550/arXiv.2503.08626 (2025)

    2025 doi

  8. [15]

    M. S. Rudolph, J. Chen, J. Miller, A. Acharya, and A. Perdomo-Ortiz, Decomposition of matrix product states into shallow quantum circuits, Quantum Science and Technology 9, 015012 (2023)

    2023

  9. [16]

    M. S. Rudolph, J. Miller, D. Motlagh, J. Chen, A. Acharya, and A. Perdomo-Ortiz, Synergistic pretrain- ing of parametrized quantum circuits via tensor net- works, Nature Communications 14, 8367 (2023)

    2023

  10. [17]

    Dborin, F

    J. Dborin, F. Barratt, V. Wimalaweera, L. Wright, and A. G. Green, Matrix product state pre-training for quan- tum machine learning, Quantum Science and Technology 7, 035014 (2022)

    2022

  11. [18]

    Jamet, C

    F. Jamet, C. Lenihan, L. P. Lindoy, A. Agarwal, E. Fontana, B. A. Martin, and I. Rungger, Anderson im- purity solver integrating tensor network methods with quantum computing, arXiv preprint arXiv:2304.06587 12 (2023)

    2023

  12. [19]

    Rogerson and A

    D. Rogerson and A. Roy, Quantum circuit optimiza- tion using differentiable programming of tensor network states, arXiv preprint arXiv:2408.12583 (2024)

    2024 arXiv

  13. [20]

    S.-H. Lin, R. Dilip, A. G. Green, A. Smith, and F. Poll- mann, Real-and imaginary-time evolution with com- pressed quantum circuits, PRX Quantum 2, 010342 (2021)

    2021

  14. [21]

    Anselme Martin, T

    B. Anselme Martin, T. Ayral, F. Jamet, M. J. Ranˇ ci´ c, and P. Simon, Combining matrix product states and noisy quantum computers for quantum simulation, Phys- ical Review A 109, 062437 (2024)

    2024

  15. [22]

    Causer, F

    L. Causer, F. Jung, A. Mitra, F. Pollmann, and A. Gammon-Smith, Scalable simulation of nonequilib- rium quantum dynamics via classically optimized unitary circuits, Physical Review Research 6, 033062 (2024)

    2024

  16. [23]

    Mc Keever and M

    C. Mc Keever and M. Lubasch, Classically optimized Hamiltonian simulation, Physical review research 5, 023146 (2023)

    2023

  17. [24]

    Gibbs and L

    J. Gibbs and L. Cincio, Deep circuit compression for quantum dynamics via tensor networks, arXiv preprint arXiv:2409.16361 (2024)

    2024 arXiv

  18. [25]

    Zhang, R

    Y. Zhang, R. Wiersema, J. Carrasquilla, L. Cincio, and Y. B. Kim, Scalable quantum dynamics compi- lation via quantum machine learning, arXiv preprint arXiv:2409.16346 (2024)

    2024 arXiv

  19. [26]

    I. N. M. Le, S. Sun, and C. B. Mendl, Riemannian quan- tum circuit optimization based on matrix product oper- ators, arXiv preprint arXiv:2501.08872 (2025)

    2025

  20. [27]

    Mansuroglu, T

    R. Mansuroglu, T. Eckstein, L. N¨ utzel, S. A. Wilkinson, and M. J. Hartmann, Variational Hamiltonian simula- tion for translational invariant systems via classical pre- processing, Quantum Science and Technology 8, 025006 (2023)

    2023

  21. [29]

    Vidal, Classical simulation of infinite-size quantum lattice systems in one spatial dimension, Physical Review Letters 98, 070201 (2007)

    G. Vidal, Classical simulation of infinite-size quantum lattice systems in one spatial dimension, Physical Review Letters 98, 070201 (2007)

    2007

  22. [30]

    Zauner-Stauber, L

    V. Zauner-Stauber, L. Vanderstraeten, M. T. Fishman, F. Verstraete, and J. Haegeman, Variational optimization algorithms for uniform matrix product states, Physical Review B 97, 045145 (2018)

    2018

  23. [31]

    M. C. Caro, H.-Y. Huang, N. Ezzell, J. Gibbs, A. T. Sornborger, L. Cincio, P. J. Coles, and Z. Holmes, Out- of-distribution generalization for learning quantum dy- namics, Nature Communications 14, 3751 (2023)

    2023

  24. [32]

    Sokolov and J

    I. Sokolov and J. Dziarmaga, Bang-bang preparation of a quantum many-body ground state in a finite lattice: op- timization of the algorithm with a tensor network, arXiv preprint arXiv:2505.08226 10.48550/arXiv.2505.08226 (2025)

    2025 doi arXiv

  25. [33]

    R. B. Lehoucq, D. C. Sorensen, and C. Yang, ARPACK users’ guide: solution of large-scale eigenvalue problems with implicitly restarted Arnoldi methods(SIAM, 1998)

    1998

  26. [34]

    Mogensen and A

    P. Mogensen and A. Riseth, Optim: A mathematical op- timization package for julia, Journal of Open Source Soft- ware 3, 10.21105/joss.00615 (2018)

    2018 doi

  27. [35]

    W. E. Thirring, A soluble relativistic field theory, Annals of Physics 3, 91 (1958)

    1958

  28. [36]

    Kogut and L

    J. Kogut and L. Susskind, Hamiltonian formulation of wilson’s lattice gauge theories, Physical Review D 11, 395 (1975)

    1975

  29. [37]

    Pfeuty, The one-dimensional Ising model with a trans- verse field, ANNALS of Physics 57, 79 (1970)

    P. Pfeuty, The one-dimensional Ising model with a trans- verse field, ANNALS of Physics 57, 79 (1970)

    1970

  30. [38]

    Schollw¨ ock, The density-matrix renormalization group in the age of matrix product states, Annals of physics 326, 96 (2011)

    U. Schollw¨ ock, The density-matrix renormalization group in the age of matrix product states, Annals of physics 326, 96 (2011)

    2011

  31. [39]

    R. R. Tucci, An Introduction to Cartan’s KAK Decom- position for QC Programmers, arXiv:quant-ph/0507171 (2005)

    2005 arXiv

  32. [40]

    N. J. Ross and P. Selinger, Optimal ancilla-free clif- ford+T approximation of z-rotations, Quantum Info. Comput. 16, 901–953 (2016)

    2016

  33. [41]

    F. Jin, D. Willsch, M. Willsch, H. Lagemann, K. Michielsen, and H. De Raedt, Random state technol- ogy, Journal of the Physical Society of Japan 90, 012001 (2021)

    2021

  34. [42]

    Learning Circuits with Infinite Tensor Networks

    T. Hao, A. Xu, and S. Tannu, Reducing T gates with unitary synthesis, arXiv preprint arXiv:2503.15843 10.48550/arXiv.2503.15843 (2025). 1 Supplementary Material for “Learning Circuits with Infinite Tensor Networks” Appendix A: Generalization in Unitary Compilation Here we stud...

    2025 doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.