Pith. sign in

REVIEW 3 major objections 8 minor 1 cited by

The paper claims that a Python library, SeeMPS, turns Matrix Product States and Tensor Trains into a finite-precision linear algebra, so standard solvers and PDE routines can run on exponentially large compressed vectors.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

SeeMPS is a Python MPS/TT library offering a BLAS/LAPACK-style API for compressed linear algebra, from DMRG and time evolution to PDE solving and Fourier transforms.

T0 review reviewed 2026-08-03 challenge →

load-bearing objection SeeMPS is a serious, well-engineered MPS/TT library whose paper overstates the validation of its 'finite-precision linear algebra' claim by omitting any benchmarks; worth a referee, but needs a revision. the 3 major comments →

arxiv 2601.16734 v2 pith:VCEFFAZS submitted 2026-01-23 quant-ph cs.NAmath.NA

SeeMPS: A Python-based Matrix Product State and Tensor Train Library

Paula Garc\'ia-Molina , Juan Jos\'e Rodr\'iguez-Aldavero , Jorge Gidi , Juan Jos\'e Garc\'ia-Ripoll This is my paper
classification quant-ph cs.NAmath.NA
keywords matrix product statestensor trainsquantized tensor trainsfinite-precision linear algebraDMRGquantum-inspired numerical analysisPDE solversPython library
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

SeeMPS is a Python library that treats Matrix Product States (MPS) and Tensor Trains (TT) as compressed stand-ins for ordinary vectors, and Matrix Product Operators (MPOs) as compressed matrices. The paper argues that with these objects and a small set of low-level operations—addition, inner products, matrix-vector products, Hadamard products—all followed by a variational compression step, standard high-level algorithms can be rebuilt: eigenvalue searches, linear-system solvers, quantum Fourier transforms, differentiation, integration, and PDE solvers. If that claim holds, exponentially large vectors become practical, first-class numerical objects, and the same code can serve quantum many-body physics and quantum-inspired numerical analysis. The library is presented as production-oriented: roughly 15,000 lines of Python and Cython, with about 93% of the production code exercised by unit tests.

Core claim

The paper's central claim is that MPS/TT is not just a compression tool but a complete finite-precision arithmetic for vectors and operators. Every basic operation returns a new MPS/MPO whose bond dimension is capped by a user-supplied strategy, with a truncation error estimate stored on the object. On top of this 'MPS-BLAS' layer, the paper implements an 'MPS-LAPACK' layer—DMRG, gradient descent, power method, restarted Arnoldi, CGS, BiCGS, GMRES, linear-system DMRG, and QFT-based Fourier transforms—and reports that conventional NumPy-style algorithms translate almost verbatim onto these objects. It then extends the same formalism to quantized tensor trains for functions, so multidimensiona

What carries the argument

The load-bearing object is the MPS/TT decomposition itself: a vector of dimension d^L is written as a trace of L small tensors with bond dimension χ, and an MPO analogously stores a linear operator. The operation that makes the algebra closed is the variational simplification routine—an iterative two-site optimization that replaces any MPS with the nearest MPS of bounded bond dimension, Eq. (10)—applied implicitly after additions, MPO-MPS products, and evolution steps. A Strategy object fixes the maximum bond dimension and tolerance, and an accumulated error field tracks the worst-case truncation introduced by all operations. This simplification step is what turns otherwise intractable bond-

Load-bearing premise

The load-bearing premise is that the variational simplification step always finds a near-optimal MPS/TT approximation of bounded bond dimension, so that truncation errors over long chains of operations stay small and can be tracked by the reported error field.

What would settle it

Take a concrete long computation—for instance, many time steps of a nonlinear or long-range Hamiltonian—and compare the SeeMPS solution against an exact or high-bond-dimension reference. If the accumulated error field stays small while the true error grows, or if the stored error is not an upper bound on the actual deviation, the finite-precision abstraction is not doing its job. A simpler version: repeatedly apply a random MPO to an exactly representable MPS, simplifying after each step, and measure the true distance to the exact result against the sum of per-step truncation estimates.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Standard numerical recipes (conjugate gradients, GMRES, Runge-Kutta, Crank-Nicolson) can be written almost unchanged, with MPS/MPO objects replacing vectors and matrices.
  • Because the same vector/operator abstraction underlies quantum states and discretized functions, a single codebase covers DMRG, TEBD, quantum-circuit emulation, and high-dimensional PDEs.
  • With QTT encoding, a function on a grid of 2^{dn} points can be stored in O(dnχ²) parameters, making high-dimensional representation practical when correlations are low.
  • The QFT MPO keeps bond dimensions nearly constant, so Fourier-based differentiation and interpolation can achieve exponential accuracy in the number of qubits.
  • The presence of an accumulated error estimate on every MPS gives algorithms a built-in, though coarse, certificate of how much the compressed representation may have drifted.

Where Pith is reading between the lines

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

  • Editorial inference: if the variational simplification reliably preserves near-optimality over long chains, SeeMPS could become a general tensor-network workhorse for high-dimensional scientific computing; the authors do not demonstrate long-horizon error stability in the paper itself.
  • Editorial inference: the error field stored on MPS objects could be promoted from a witness to a control signal, letting adaptive time-stepping or solver truncation react to measured deterioration—a use the paper sketches but does not develop.
  • Editorial inference: the polynomial-expansion machinery for operator-valued functions suggests immediate applications beyond PDEs, such as fractional derivatives, matrix exponentials, and spectral filters, without new infrastructure.
  • Editorial inference: the claim of complete coverage would be strengthened by independent cross-checks against existing tensor-network software on shared benchmarks; the paper provides internal tests but no such comparison.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 8 minor

Summary. The manuscript introduces SeeMPS, a Python/Cython library for matrix product states (MPS) and tensor trains (TT), including matrix product operators (MPOs). It presents the library as a finite-precision linear algebra package: MPS/TT vectors and MPO operators support scaling, addition, inner products, matrix-vector products, Hadamard products, and tensor products, with implicit simplification and error estimates; higher-level solvers (DMRG, power/Arnoldi, CGS/GMRES, Fourier transforms), function encoding/interpolation/differentiation/integration, static and time-dependent PDE solvers, and quantum-circuit emulation are built on these primitives. The paper's main claim is that this abstraction makes high-level MPS/TT algorithms practical and reliable for quantum many-body and quantum-inspired numerical analysis.

Significance. SeeMPS addresses a genuine gap: most existing tensor-network software is domain-specific, while a unified, BLAS/LAPACK-style MPS/TT library would ease adoption. The engineering strengths are real and should be acknowledged: open-source MIT-licensed code, a Cython-optimized core, 93% unit-test coverage, CI/CD with type checking and linting, and online documentation. However, the present manuscript contains no numerical experiments, convergence studies, or performance benchmarks. Since the central claim is that the library's operations have controlled truncation error and that high-level algorithms are reliable, this missing validation is decisive for the paper's evaluation.

major comments (3)
  1. [Sect. 2.3, Eq. (10)] The central abstraction of the package as a finite-precision linear algebra package (Abstract; Sect. 2.4) rests on the default variational simplification (Eq. 10) returning a near-optimal MPS approximation with a reliable error estimate after each operation. This assumption is not established: Eq. (10) is a nonconvex optimization over MPS_chi, solved by alternating local updates, and the reported error is only a 'crude witness of possible deterioration' (Sect. 2.3), not a bound on ||psi_opt - phi||. The manuscript contains no numerical experiment showing that accumulated errors stay controlled in long chains of operations, e.g., in CGS (Fig. 4), Arnoldi (Sect. 3.1.4), or time evolution (Sect. 7). A benchmark comparing exact or high-rank reference solutions with the error tracked by the library is needed to support the central claim.
  2. [Sects. 3-8 (overall)] Throughout Sections 3-8 the text makes quantitative claims without a single numerical table or figure. For example, the abstract and Sect. 3.3 claim 'exponentially efficient Fourier transforms'; Sect. 3.2.3 says algorithms 'rapidly converge'; Sect. 9 asserts 'competitive performance'. The reader cannot verify that the implemented algorithms work on nontrivial sizes, that bond dimensions remain controlled, or that the claimed runtime scalings hold. For a software paper, at least one reproducible example per solver family with reported errors, timings, and configurations is standard; its absence leaves the main utility claims unsubstantiated.
  3. [Sect. 9, Table 2] The 93% unit-test coverage figure and CI setup are commendable, but they do not establish the mathematical claim of controlled finite-precision error. Coverage quantifies how many code lines are executed by tests, not whether the default simplification (Eq. 10) maintains accuracy over long operation sequences or whether iterative solvers converge on realistic problem sizes. The paper states that tests enforce that 'algorithms satisfy predefined error bounds', but no test results are shown. Either include the actual checks and error measurements, or soften the claim that this constitutes evidence for the finite-precision abstraction.
minor comments (8)
  1. [Eq. (4)] The summation index i is reused for the vector index and the dummy variable; use a different index in the formula for the tensorization map.
  2. [Eq. (6)] The first core A^(1) is written with an alpha_L index, which is inconsistent with the boundary convention alpha_L = 1 described in the same section; check the trace formula and indexing.
  3. [Eq. (28)] The symbol v_m is used both as a scalar coefficient and as the vector v; rename the coefficients (e.g., c_m) to avoid confusion.
  4. [Eq. (82)] The symmetric Trotter decomposition should presumably read U_0(dt/2) U_1(dt) U_0(dt/2); the displayed third factor lacks the 1/2.
  5. [Eq. (75)] The symbol hbar in the Radau integration formula is undefined; if it is a scaling parameter, define it, otherwise remove it.
  6. [Sect. 3.3, Eqs. (33)-(34), (52)] Equation numbers are out of sequence and the QFT tensor indices in Eq. (52) are ambiguous; renumber and clarify the convention.
  7. [References] References [45] and [46] are Wikipedia articles; for a journal submission, standard numerical-analysis textbooks or original papers would be more appropriate.
  8. [General] There are numerous typographical issues ('nearest-neighbourgh', 'as sketeched', 'c.f.'), and the text would benefit from a careful copyedit.

Circularity Check

0 steps flagged

No circular derivation found: SeeMPS is an implementation report whose algorithms are standard or externally supported; the identified weaknesses are evidentiary gaps, not circular reductions.

full rationale

The paper is a software and methods description rather than a derivation of new predictions. Its core equations are definitions and standard optimization problems: the MPS decomposition (Eq. 6), MPO decomposition (Eq. 8), matrix-vector contraction (Eq. 12), and the variational simplifications (Eqs. 10, 11, 13) formalize what it means to project onto a bounded-bond-dimension manifold. No quantity is fitted to a subset of data and then relabeled as a prediction; no uniqueness theorem is imported from the authors' prior work to force an ansatz; and no known empirical pattern is renamed as a new theory. The heavy self-citation ([8], [13], [19], [26], [49]) is noticeable, but it is not load-bearing in a circular way: [8] describes the variational MPS compression algorithm from 2006, [19] contains earlier comparisons of the finite-precision algebra, and the paper also cites many external, independent sources for DMRG, TT, TCI, polynomial expansions, Krylov methods, and TEBD (e.g., [1,2,3,15,16,29,37,45,46,47,67,68]). The paper explicitly flags its own limitation: the accumulated errors are stored 'as a crude witness of possible deterioration of the approximation' and admits that the estimate is not a proven tight bound ('Though in practice the results may be better than the estimates [8]'). This is an honest acknowledgment of an evaluation gap, not a circular step. The lack of end-to-end error-accumulation benchmarks for long chains of operations is a correctness/evidence concern and should be assessed under correctness risk, not circularity. Therefore the appropriate finding is no significant circularity.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

The paper contains no fitted scientific parameters, no new physical entities, and no free parameters beyond user-specified tolerances/bond dimensions. The axioms listed are the background assumptions from numerical analysis and prior literature that the library's efficiency claims rest on.

axioms (4)
  • domain assumption Bond dimensions of relevant MPS/QTT representations remain small for the target problem classes (smooth functions, low-entanglement states).
    All memory and runtime claims depend on this; the paper cites Refs [16,17,50] for support but does not verify it for the library's PDE/sampling applications.
  • domain assumption The variational compression in Eq. (10) converges to a near-optimal approximation in all operations, with controlled truncation error.
    Section 2.3 describes this as the default simplification; the paper itself calls the error estimate a 'crude witness' and notes results 'may be better than the estimates [8]'.
  • standard math The Quantum Fourier Transform MPO has small bond dimension (Chen, Stoudenmire, White, PRX Quantum 4, 040318).
    Sect. 3.3 relies on this to claim exponentially efficient Fourier transforms; it is an external theorem, not proved here.
  • domain assumption Tensor cross-interpolation maxvol heuristics select informative pivots, so TCI-based encodings are accurate.
    Section 4.5 uses TCI as a default loader; the paper itself notes DMRG-based TCI 'can be more susceptible to convergence to local minima', so accuracy is not guaranteed.

reviewed 2026-08-03 · how reviews work

0 comments
Cite this review

Pith. "Pith review of SeeMPS: A Python-based Matrix Product State and Tensor Train Library." pith.science (2026). https://pith.science/paper/VCEFFAZS

@misc{pith2026260116734,
  author       = {Pith},
  title        = {Pith review of: SeeMPS: A Python-based Matrix Product State and Tensor Train Library},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VCEFFAZS}},
  note         = {Machine review of arXiv:2601.16734}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We introduce SeeMPS, a Python library dedicated to implementing tensor network algorithms based on the well-known Matrix Product States (MPS) and Quantized Tensor Train (QTT) formalisms. SeeMPS is implemented as a complete finite precision linear algebra package where exponentially large vector spaces are compressed using the MPS/TT formalism. It enables both low-level operations, such as vector addition, linear transformations, and Hadamard products, as well as high-level algorithms, including the approximation of linear equations, eigenvalue computations, and exponentially efficient Fourier transforms. This library can be used for traditional quantum many-body physics applications and also for quantum-inspired numerical analysis problems, such as solving PDEs, interpolating and integrating multidimensional functions, sampling multivariate probability distributions, etc.

Figures

Figures reproduced from arXiv: 2601.16734 by Jorge Gidi, Juan Jos\'e Garc\'ia-Ripoll, Juan Jos\'e Rodr\'iguez-Aldavero, Paula Garc\'ia-Molina.

Figure 1
Figure 1. Figure 1: (a) A high-dimensional vector vi is reinterpreted as a tensor vs1,s2,...,sN that can be decomposed in tensor train or matrix product state (MPS) forms, with smaller tensors A (n),in αn,αn+1 that occupy less space. (b) A similar procedure applied to matrices leads to what is known as a matrix product operator (MPO). of the library’s architecture in Sect. 9 and some general conclusions and outlook in Sect. 1… view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of core MPS-BLAS operations in SeeMPS: (a) [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The tensor product of two vectors v⊗w represented by tensor trains {A} and {B} is constructed in SeeMPS using (a) a sequential “A” or an (b) interleaved “B” order. an operator onto one or two vectors: e.g., Q.expectation(u,v) produces ⟨u, Qv⟩. 2.4.5. Element-wise products Motivated by applications in the study of nonlinear equations, SeeMPS also offers the element-wise product or Hadamard product of two ve… view at source ↗
Figure 4
Figure 4. Figure 4: Pseudocode implementation of the conjugate gradient solver [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Discretization and encoding of a periodic function on a grid [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Pseudocode to encode a multivariate function [PITH_FULL_IMAGE:figures/full_fig_p013_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Pseudocode to encode a multivariate function [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Pseudocode to create a QAOA model for an Ising Hamilto [PITH_FULL_IMAGE:figures/full_fig_p019_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: SeeMPS’s architecture. Cython code that provides optimized functions to (i) de￾compose a tensor into two tensors using different trunca￾tion strategies, (ii) implement key multidimensional ten￾sor contractions used in various algorithms, (iii) compute an MPS’s canonical form, and (iv) implement a scalar product among MPS. The choice of which components to optimize is based on a careful study of performance… view at source ↗

discussion (0)

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

Forward citations

Cited by 1 Pith paper

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

  1. Performance Benchmarking: Software for the Density Matrix Renormalization Group

    physics.comp-ph 2026-07 accept novelty 6.0

    A cost-and-quality benchmarking framework applied to eight DMRG codes finds up to 100× performance gaps between packages and between parameter settings within a package.

Reference graph

Works this paper leans on

79 extracted references · 17 canonical work pages · cited by 1 Pith paper · 1 internal anchor

  1. [1]

    S. R. White, Density matrix formulation for quan- tum renormalization groups, Phys. Rev. Lett. 69 (19) (1992) 2863–2866.doi:10.1103/PhysRevLett. 69.2863

    doi:10.1103/physrevlett 1992

  2. [2]

    S. R. White, Density-matrix algorithms for quantum renormalization groups, Phys. Rev. B 48 (14) (1993) 10345–10356.doi:10.1103/PhysRevB.48.1034 5

    work page doi:10.1103/physrevb.48.1034 1993

  3. [3]

    Schollwöck, The density-matrix renormalization group in the age of matrix product states, Annals of Physics 326 (1) (2011) 96–192.doi:10.1016/j

    U. Schollwöck, The density-matrix renormalization group in the age of matrix product states, Annals of Physics 326 (1) (2011) 96–192.doi:10.1016/j. aop.2010.09.012

    doi:10.1016/j 2011

  4. [4]

    Verstraete, D

    F. Verstraete, D. Porras, J. I. Cirac, Density Ma- trix Renormalization Group and Periodic Bound- ary Conditions: A Quantum Information Perspec- tive, Phys. Rev. Lett. 93 (22) (2004) 227205.doi: 10.1103/PhysRevLett.93.227205

    doi:10.1103/physrevlett.93.227205 2004

  5. [5]

    Porras, F

    D. Porras, F. Verstraete, J. I. Cirac, Renormalization algorithm for the calculation of spectra of interact- ing quantum systems, Phys. Rev. B 73 (1) (2006) 014410.doi:10.1103/PhysRevB.73.014410

    work page doi:10.1103/physrevb.73.014410 2006

  6. [8]

    J. J. García-Ripoll, Time evolution of Matrix Product States, New J. Phys. 8 (12) (2006) 305.doi:10.1 088/1367-2630/8/12/305

    2006

  7. [10]

    J. I. Latorre, Image compression and entanglement (Oct. 2005).arXiv:quant-ph/0510031,doi: 10.48550/arXiv.quant-ph/0510031

    Pith/arXiv arXiv doi:10.48550/arxiv.quant-ph/0510031 2005

  8. [13]

    J. J. García-Ripoll, Quantum-inspired algorithms for multivariate analysis: From interpolation to partial differential equations, Quantum 5 (2021) 431.doi: 10.22331/q-2021-04-15-431

    doi:10.22331/q-2021-04-15-431 2021

  9. [14]

    Gourianov, M

    N. Gourianov, M. Lubasch, S. Dolgov, Q. Y . van den Berg, H. Babaee, P. Givi, M. Kiffner, D. Jaksch, A quantum-inspired approach to exploit turbulence structures, Nat Comput Sci 2 (1) (2022) 30–37.doi: 10.1038/s43588-021-00181-1

    doi:10.1038/s43588-021-00181-1 2022

  10. [15]

    I. V . Oseledets, Tensor-Train Decomposition, SIAM Journal on Scientific Computingdoi:10.1137/09 0752286

    work page doi:10.1137/09

  11. [16]

    I. V . Oseledets, Approximation of $2^d\times2^d$ Matrices Using Tensor Decomposition, SIAM Jour- nal on Matrix Analysis and Applicationsdoi:10.1 137/090757861

  12. [17]

    B. N. Khoromskij, I. V . Oseledets, QTT approxima- tion of elliptic solution operators in higher dimen- sions 26 (3) (2011) 303–322.doi:10.1515/rjna mm.2011.017

    work page doi:10.1515/rjna 2011

  13. [18]

    B. N. Khoromskij, O(dlog N)-Quantics Approxima- tion of N-d Tensors in High-Dimensional Numerical Modeling, Constr Approx 34 (2) (2011) 257–280. doi:10.1007/s00365-011-9131-1

    doi:10.1007/s00365-011-9131-1 2011

  14. [19]

    García-Molina, L

    P. García-Molina, L. Tagliacozzo, J. J. García- Ripoll, Global optimization of MPS in quantum- inspired numerical analysis (May 2024).arXiv:23 03.09430,doi:10.48550/arXiv.2303.09430. 22

    doi:10.48550/arxiv.2303.09430 2024

  15. [20]

    Zalka, Simulating quantum systems on a quantum computer, Proc

    C. Zalka, Simulating quantum systems on a quantum computer, Proc. A 454 (1969) (1998) 313–322.doi: 10.1098/rspa.1998.0162

    arXiv 1969

  16. [21]

    Grover, T

    L. Grover, T. Rudolph, Creating superpositions that correspond to efficiently integrable probability dis- tributions (Aug. 2002).arXiv:quant-ph/020811 2,doi:10.48550/arXiv.quant-ph/0208112

    Pith/arXiv arXiv doi:10.48550/arxiv.quant-ph/0208112 2002

  17. [22]

    B. N. Khoromskij, I. Oseledets, Quantics-TT Collo- cation Approximation of Parameter-Dependent and Stochastic Elliptic PDEs, Computational Methods in Applied Mathematics 10 (4) (2010) 376–394.doi: 10.2478/cmam-2010-0023

    doi:10.2478/cmam-2010-0023 2010

  18. [24]

    J. C. Halimeh, F. Kolley, I. P. McCulloch, Cheby- shev matrix product state approach for time evolu- tion, Phys. Rev. B 92 (11) (2015) 115130.doi: 10.1103/PhysRevB.92.115130

    work page doi:10.1103/physrevb.92.115130 2015

  19. [25]

    Lindsey, Multiscale interpolative construction of quantized tensor trains (Apr

    M. Lindsey, Multiscale interpolative construction of quantized tensor trains (Apr. 2024).arXiv:2311.1 2554,doi:10.48550/arXiv.2311.12554

    Pith/arXiv arXiv doi:10.48550/arxiv.2311.12554 2024

  20. [26]

    J. J. Rodríguez-Aldavero, P. García-Molina, L. Tagliacozzo, J. J. García-Ripoll, Chebyshev approximation and composition of functions in matrix product states for quantum-inspired numer- ical analysis (Feb. 2025).arXiv:2407.09609, doi:10.48550/arXiv.2407.09609

    Pith/arXiv arXiv doi:10.48550/arxiv.2407.09609 2025

  21. [27]

    J. Chen, E. Stoudenmire, S. R. White, Quantum Fourier Transform Has Small Entanglement, PRX Quantum 4 (4) (2023) 040318.doi:10.1103/PR XQuantum.4.040318

    doi:10.1103/pr 2023

  22. [28]

    García-Molina, J

    P. García-Molina, J. Rodríguez-Mediavilla, J. J. García-Ripoll, Quantum Fourier analysis for mul- tivariate functions and applications to a class of Schrödinger-type partial differential equations, Phys. Rev. A 105 (1) (2022) 012433.doi:10.1 103/PhysRevA.105.012433

    2022

  23. [29]

    Oseledets, E

    I. Oseledets, E. Tyrtyshnikov, TT-cross approxima- tion for multidimensional arrays, Linear Algebra and its Applications 432 (1) (2010) 70–88.doi:10.101 6/j.laa.2009.07.024

    2010

  24. [30]

    Savostyanov, I

    D. Savostyanov, I. Oseledets, Fast adaptive interpo- lation of multi-dimensional arrays in tensor train for- mat, in: The 2011 International Workshop on Mul- tidimensional (nD) Systems, 2011, pp. 1–8.doi: 10.1109/nDS.2011.6076873

    arXiv 2011

  25. [31]

    Mikhalev, I

    A. Mikhalev, I. Oseledets, Rectangular maximum- volume submatrices and their applications, Linear Algebra and its Applications 538 (2018) 187–211. doi:10.1016/j.laa.2017.10.014

    doi:10.1016/j.laa.2017.10.014 2018

  26. [32]

    Dolgov, D

    S. Dolgov, D. Savostyanov, Parallel cross inter- polation for high-precision calculation of high- dimensional integrals, Computer Physics Communi- cations 246 (2020) 106869.doi:10.1016/j.cpc. 2019.106869

    arXiv 2020

  27. [33]

    Núñez Fernández, M

    Y . Núñez Fernández, M. Jeannin, P. T. Dumitrescu, T. Kloss, J. Kaye, O. Parcollet, X. Waintal, Learning Feynman Diagrams with Tensor Trains, Phys. Rev. X 12 (4) (2022) 041018.doi:10.1103/PhysRevX .12.041018

    doi:10.1103/physrevx 2022

  28. [34]

    Núñez Fernández, M

    Y . Núñez Fernández, M. K. Ritter, M. Jeannin, J.- W. Li, T. Kloss, T. Louvet, S. Terasaki, O. Parcollet, J. V on Delft, H. Shinaoka, X. Waintal, Learning ten- sor networks with tensor cross interpolation: New algorithms and libraries, SciPost Phys. 18 (3) (2025) 104.doi:10.21468/SciPostPhys.18.3.104

    doi:10.21468/scipostphys.18.3.104 2025

  29. [35]

    Sozykin, A

    K. Sozykin, A. Chertkov, R. Schutski, A.-H. Phan, A. S. Cichocki, I. Oseledets, TTOpt: A Maximum V olume Quantized Tensor Train-based Optimization and its Application to Reinforcement Learning, Ad- vances in Neural Information Processing Systems 35 (2022) 26052–26065

    2022

  30. [36]

    M. K. Ritter, Y . Núñez Fernández, M. Wallerberger, J. V on Delft, H. Shinaoka, X. Waintal, Quantics Ten- sor Cross Interpolation for High-Resolution Parsi- monious Representations of Multivariate Functions, Phys. Rev. Lett. 132 (5) (2024) 056501.doi: 10.1103/PhysRevLett.132.056501

    doi:10.1103/physrevlett.132.056501 2024

  31. [37]

    J. I. Cirac, D. Pérez-García, N. Schuch, F. Verstraete, Matrix product states and projected entangled pair states: Concepts, symmetries, theorems, Rev. Mod. Phys. 93 (4) (2021) 045003.doi:10.1103/RevMod Phys.93.045003

    doi:10.1103/revmod 2021

  32. [38]

    Schollwöck, The density-matrix renormalization group, Rev

    U. Schollwöck, The density-matrix renormalization group, Rev. Mod. Phys. 77 (1) (2005) 259–315.do i:10.1103/RevModPhys.77.259. 23

    doi:10.1103/revmodphys.77.259 2005

  33. [40]

    Dukelsky, M

    J. Dukelsky, M. A. Martín-Delgado, T. Nishino, G. Sierra, Equivalence of the variational matrix product method and the density matrix renormaliza- tion group applied to spin chains, EPL 43 (4) (1998) 457.doi:10.1209/epl/i1998-00381-x

    work page doi:10.1209/epl/i1998-00381-x 1998

  34. [41]

    A. A. Gorodetsky, J. D. Jakeman, Gradient-based optimization for regression in the functional tensor- train format, Journal of Computational Physics 374 (2018) 1219–1238.doi:10.1016/j.jcp.2018.0 8.010

    work page doi:10.1016/j.jcp.2018.0 2018

  35. [42]

    J. Wang, C. Roberts, G. Vidal, S. Leichenauer, Anomaly detection with tensor networks, arXiv e- prints arXiv:2006.02516.doi:10.48550/ARXIV.2 006.02516

    Pith/arXiv arXiv doi:10.48550/arxiv.2006.02516 2006

  36. [43]

    Barratt, J

    F. Barratt, J. Dborin, L. Wright, Improvements to gradient descent methods for quantum ten- sor network machine learning, arXiv e-prints arXiv:2203.03366.doi:10.48550/ARXIV.220 3.03366

    Pith/arXiv arXiv doi:10.48550/arxiv.2203.03366

  37. [44]

    Pollock, L

    S. Pollock, L. Scott, Extrapolating the arnoldi algo- rithm to improve eigenvector convergence, Interna- tional Journal of Numerical Analysis and Modeling 18 (5) (2021) 712–721

    2021

  38. [45]

    Conjugate gradient method, Wikipedia

  39. [46]

    Biconjugate gradient method, Wikipedia

  40. [47]

    T. D. Kühner, S. R. White, Dynamical correlation functions using the density matrix renormalization group, Phys. Rev. B 60 (1) (1999) 335–343.doi: 10.1103/PhysRevB.60.335

    work page doi:10.1103/physrevb.60.335 1999

  41. [48]

    Coppersmith, An approximate Fourier transform useful in quantum factoring (Jan

    D. Coppersmith, An approximate Fourier transform useful in quantum factoring (Jan. 2002).arXiv: quant-ph/0201067,doi:10.48550/arXiv.qua nt-ph/0201067

    Pith/arXiv arXiv doi:10.48550/arxiv.quant-ph/0201067 2002

  42. [49]

    J. Gidi, P. García-Molina, L. Tagliacozzo, J. J. García-Ripoll, Pseudospectral method for solving PDEs using matrix product states, Journal of Com- putational Physics 539 (2025) 114228.doi:10.101 6/j.jcp.2025.114228

    arXiv 2025

  43. [50]

    Jobst, K

    B. Jobst, K. Shen, C. A. Riofrío, E. Shishen- ina, F. Pollmann, Efficient MPS representations and quantum circuits from the Fourier modes of clas- sical image data, Quantum 8 (2024) 1544.doi: 10.22331/q-2024-12-03-1544

    work page doi:10.22331/q-2024-12-03-1544 2024

  44. [51]

    I. V . Oseledets, Constructive Representation of Functions in Low-Rank Tensor Formats, Constr Ap- prox 37 (1) (2013) 1–18.doi:10.1007/s00365-0 12-9175-x

    work page doi:10.1007/s00365-0 2013

  45. [52]

    C. W. Clenshaw, A note on the summation of Cheby- shev series, Math. Comp. 9 (51) (1955) 118–120. doi:10.1090/S0025-5718-1955-0071856-0

    doi:10.1090/s0025-5718-1955-0071856-0 1955

  46. [53]

    Driscoll, N

    T. Driscoll, N. Hale, L. N. Trefethen, Chebfun Guide, Pafnuty Publications, Oxford, 2014

    2014

  47. [54]

    D. M. Monro, Interpolation by fast fourier and chebyshev transforms, International Journal for Nu- merical Methods in Engineering 14 (11) (1979) 1679–1692.arXiv:https://onlinelibrary.wi ley.com/doi/pdf/10.1002/nme.1620141109, doi:10.1002/nme.1620141109

    work page doi:10.1002/nme.1620141109 1979

  48. [55]

    Nyquist, Certain Topics in Telegraph Transmis- sion Theory, Transactions of the American Insti- tute of Electrical Engineers 47 (2) (1928) 617–644

    H. Nyquist, Certain Topics in Telegraph Transmis- sion Theory, Transactions of the American Insti- tute of Electrical Engineers 47 (2) (1928) 617–644. doi:10.1109/t-aiee.1928.5055024

    arXiv 1928

  49. [56]

    C. E. Shannon, Communication in the Presence of Noise, Proceedings of the IRE 37 (1) (1949) 10–21. doi:10.1109/jrproc.1949.232969

    arXiv 1949

  50. [57]

    Goreinov, E

    S. Goreinov, E. Tyrtyshnikov, N. Zamarashkin, A theory of pseudoskeleton approximations, Linear Algebra and its Applications 261 (1-3) (1997) 1–21. doi:10.1016/S0024-3795(96)00301-1

    work page doi:10.1016/s0024-3795(96)00301-1 1997

  51. [58]

    S. A. Goreinov, I. V . Oseledets, D. V . Savostyanov, E. E. Tyrtyshnikov, N. L. Zamarashkin, How to Find a Good Submatrix, WORLD SCIENTIFIC, 2010, pp. 247–256.doi:10.1142/9789812836021_ 0015

    work page doi:10.1142/9789812836021_ 2010

  52. [59]

    Bigoni, A

    D. Bigoni, A. P. Engsig-Karup, Y . M. Marzouk, Spectral Tensor-Train Decomposition, SIAM J. Sci. Comput. 38 (4) (2016) A2405–A2439.doi:10.113 7/15M1036919

    2016

  53. [60]

    D. Adak, M. E. Danis, D. P. Truong, K. Ø. Ras- mussen, B. S. Alexandrov, Tensor Network Space- Time Spectral Collocation Method for Solving the Nonlinear Convection Diffusion Equation, J Sci 24 Comput 103 (2) (2025) 46.doi:10.1007/s109 15-025-02860-x

    doi:10.1007/s109 2025

  54. [61]

    J. Chen, M. Lindsey, Direct interpolative construc- tion of the discrete Fourier transform as a matrix product operator, Applied and Computational Har- monic Analysis 81 (2026) 101817.doi:10.1016/ j.acha.2025.101817

    arXiv 2026

  55. [62]

    Y . Hur, J. G. Hoskins, M. Lindsey, E. Stouden- mire, Y . Khoo, Generative modeling via tensor train sketching, Applied and Computational Harmonic Analysis 67 (2023) 101575.doi:10.1016/j.ac ha.2023.101575

    arXiv 2023

  56. [63]

    J. R. Pareja Monturiol, A. Pozas-Kerstjens, D. Pérez-García, Tensorization of neural net- works for improved privacy and interpretabil- ity, SciPost Physics Core 8 (4) (2025) 095. doi:10.21468/SciPostPhysCore.8.4.095

    doi:10.21468/scipostphyscore.8.4.095 2025

  57. [64]

    Constructive TT-representation of the tensors given as index interaction functions with applications

    G. Ryzhakov, I. Oseledets, Constructive TT- representation of the tensors given as index interac- tion functions with applications (Jun. 2022).arXiv: 2206.03832,doi:10.48550/arXiv.2206.0383 2

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2206.03832 2022

  58. [65]

    Holoborodko, Smooth noise-robust differentiators (2008)

    P. Holoborodko, Smooth noise-robust differentiators (2008)

    2008

  59. [66]

    D. K. Hoffman, N. Nayar, O. A. Sharafeddin, D. J. Kouri, Analytic banded approximation for the dis- cretized free propagator, J. Phys. Chem. 95 (21) (1991) 8299–8305.doi:10.1021/j100174a052

    work page doi:10.1021/j100174a052 1991

  60. [67]

    Haegeman, C

    J. Haegeman, C. Lubich, I. Oseledets, B. Vanderey- cken, F. Verstraete, Unifying time evolution and op- timization with matrix product states, Phys. Rev. B 94 (16) (2016) 165116.doi:10.1103/PhysRevB .94.165116

    doi:10.1103/physrevb 2016

  61. [68]

    Vidal, Efficient Classical Simulation of Slightly Entangled Quantum Computations, Phys

    G. Vidal, Efficient Classical Simulation of Slightly Entangled Quantum Computations, Phys. Rev. Lett. 91 (14) (2003) 147902.doi:10.1103/PhysRevL ett.91.147902

    doi:10.1103/physrevl 2003

  62. [69]

    Díez-Valle, D

    P. Díez-Valle, D. Porras, J. J. García-Ripoll, Quan- tum Approximate Optimization Algorithm Pseudo- Boltzmann States, Phys. Rev. Lett. 130 (5) (2023) 050601.doi:10.1103/PhysRevLett.130.0506 01

    work page doi:10.1103/physrevlett.130.0506 2023

  63. [70]

    Díez-Valle, F

    P. Díez-Valle, F. J. Gómez-Ruiz, D. Porras, J. J. García-Ripoll, Universal Resources for QAOA and Quantum Annealing (Jun. 2025).arXiv:2506.032 41,doi:10.48550/arXiv.2506.03241

    doi:10.48550/arxiv.2506.03241 2025

  64. [71]

    Bharti, A

    K. Bharti, A. Cervera-Lierta, T. H. Kyaw, T. Haug, S. Alperin-Lea, A. Anand, M. Degroote, H. Heimo- nen, J. S. Kottmann, T. Menke, W.-K. Mok, S. Sim, L.-C. Kwek, A. Aspuru-Guzik, Noisy intermediate- scale quantum algorithms, Rev. Mod. Phys. 94 (1) (2022) 015004.doi:10.1103/RevModPhys.94. 015004

    doi:10.1103/revmodphys.94 2022

  65. [72]

    Tilly, H

    J. Tilly, H. Chen, S. Cao, D. Picozzi, K. Setia, Y . Li, E. Grant, L. Wossnig, I. Rungger, G. H. Booth, J. Tennyson, The Variational Quantum Eigensolver: A review of methods and best practices, Physics Re- ports 986 (2022) 1–128.doi:10.1016/j.physre p.2022.08.003

    work page doi:10.1016/j.physre 2022

  66. [75]

    Basedpyright Static Type Checker

  67. [76]

    Ruff, an extremely fast Python linter and code for- matter

  68. [78]

    Feiguin, J

    A. Feiguin, J. J. García-Ripoll, A. González-Tudela, Qubit-photon corner states in all dimensions, Phys. Rev. Res. 2 (2) (2020) 023082.doi:10.1103/Phys RevResearch.2.023082

    doi:10.1103/phys 2020

  69. [79]

    Fishman, S

    M. Fishman, S. White, E. M. Stoudenmire, The ITensor Software Library for Tensor Network Calcu- lations, SciPost Physics Codebases (2022) 004doi: 10.21468/SciPostPhysCodeb.4

    doi:10.21468/scipostphyscodeb.4 2022

  70. [80]

    Barcelona Supercomputing Center, Tenet.jl: A julia library for tensor networks,https://bsc-quant ic.github.io/Tenet.jl

  71. [81]

    D. I. Lyakh, T. Nguyen, D. Claudino, E. Dumitrescu, A. J. McCaskey, Exatn: Scalable gpu-accelerated high-performance processing of general tensor net- works at exascale, Frontiers in Applied Mathematics and Statistics 8 (2022) 838601

    2022

  72. [82]

    Roberts, A

    C. Roberts, A. Milsted, M. Ganahl, A. Zalcman, B. Fontaine, Y . Zou, J. Hidary, G. Vidal, S. Le- ichenauer, Tensornetwork: A library for physics and machine learning (2019).arXiv:1905.01330. 25

    Pith/arXiv arXiv 2019

  73. [83]

    Gray, quimb: a python library for quantum in- formation and many-body calculations, Journal of Open Source Software 3 (29) (2018) 819.doi: 10.21105/joss.00819

    J. Gray, quimb: a python library for quantum in- formation and many-body calculations, Journal of Open Source Software 3 (29) (2018) 819.doi: 10.21105/joss.00819

    doi:10.21105/joss.00819 2018

  74. [84]

    Hauschild, F

    J. Hauschild, F. Pollmann, Efficient numerical simu- lations with tensor networks: Tensor network python (tenpy), SciPost Physics Lecture Notes (2018) 005

    2018

  75. [85]

    Van Damme, L

    M. Van Damme, L. Devos, J. Haegeman, Mpskit.jl, Zenodo (2025).doi:10.5281/zenodo.10654901

    doi:10.5281/zenodo.10654901 2025

  76. [86]

    Oseledets, Tt-toolbox,https://github.com/o seledets/TT-Toolbox, gitHub repository (2026)

    I. Oseledets, Tt-toolbox,https://github.com/o seledets/TT-Toolbox, gitHub repository (2026)

    2026

  77. [87]

    Chertkov, G

    A. Chertkov, G. Ryzhakov, I. Oseledets, Black box approximation in the tensor train format initialized by anova decomposition, SIAM Journal on Scien- tific Computing 45 (4) (2023) A2101–A2118

    2023

  78. [88]

    J. R. P. Monturiol, D. Pérez-García, A. Pozas- Kerstjens, Tensorkrowch: Smooth integration of ten- sor networks in machine learning, Quantum 8 (2024) 1364

    2024

  79. [89]

    Sehlstedt, J

    P. Sehlstedt, J. Brandejs, P. Bientinesi, L. Karlsson, The software landscape for the density matrix renor- malization group, arXiv preprint arXiv:2506.12629. 26

    arXiv

This paper was first reviewed by deepseek-v4-flash on August 3, 2026.