Applying the Triad network representation to four-dimensional ATRG method

Shoichi Sasaki; Yuto Sugimoto

REVIEW 5 major objections 7 minor 15 references

Applying the Triad network representation to four-dimensional ATRG method

T0 review · 5 major / 7 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper proposes Triad-ATRG, a triad-network version of the four-dimensional ATRG that cuts the dominant contraction cost from O(χ^9) to O(r^2 χ^7) while matching ATRG accuracy on the four-dimensional Ising model.

desk verdict Useful, honest cost-reduction recipe for 4D ATRG, but the accuracy and GPU-scaling claims need to be tightened before I'd take them at face value. read the letter →

arxiv 2412.14104 v2 pith:Y4YQXF73 submitted 2024-12-18 hep-lat

Yuto Sugimoto , Shoichi Sasaki This is my paper · ORCID

classification hep-lat PACS 05.10.-a11.15.Ha

keywords tensorrenormalizationgroupanisotropicTRGtriadnetworkfour-dimensionalIsingmodeloversamplingrandomizedSVDGPUparallelizationfreeenergy

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

The paper sets out to make the anisotropic tensor renormalization group (ATRG) usable at large bond dimensions in four dimensions, where its O($χ^{9}$) contraction step currently blocks progress. It does so by giving the ATRG a triad network representation, adapted from the minimally-decomposed TRG, in which the unit-cell tensor is split into mostly three-leg tensors with an oversampling parameter r. The resulting Triad-ATRG has contraction cost O($r^{2}$ $χ^{7}$). Tested on the four-dimensional Ising model, it reproduces ATRG free energies to 0.0013% at χ = 54 and the phase transition point within 0.1%, while running significantly faster on both CPUs and GPUs. A sympathetic reader would care because this appears to remove the main computational obstacle to pushing four-dimensional tensor network calculations to larger bond dimensions.

What carries the argument

The triad representation of the ATRG: the unit-cell tensor Γ is decomposed by oversampled isometries into a network of 4-leg tensors (E, F, G, H) and 3-leg tensors (I, J, K, L), each dotted line carrying an oversampled bond of size rχ. This network is the load-bearing object; it converts the ATRG's single expensive O($χ^{9}$) contraction into contractions over smaller tensor products, with the oversampling parameter r controlling the trade-off between cost and truncation error. The cost count O($r^{2}$ $χ^{7}$) is what the whole argument rests on, and the numerical tests demonstrate that r = 7 preserves ATRG-level accuracy for the Ising free energy and transition temperature.

What would settle it

Compute the four-dimensional Ising free energy with Triad-ATRG at χ = 54 using r = 3 and r = 7; if the r = 3 result differs from the r = 7 result by far more than 0.0013%, or if at fixed r = 7 the deviation from ATRG grows with χ instead of shrinking, then the truncation is not uniformly controlled. A stronger test would replace the Ising model by a four-dimensional $φ^{4}$ or gauge theory where ATRG results are available and check whether the r = 7 accuracy guarantee survives.

Watch

Extended reading notes

Core claim

The central claim is that a triad representation can be grafted onto the ATRG without sacrificing its accuracy. Starting from the post-bond-swapping unit cell Γ = AXσY D, the authors multiply by oversampled isometries U_A, U_X, U_Y, U_D and contract to form a network of four-leg tensors E, F, G, H and three-leg tensors I, J, K, L. This network is then squeezed by improved isometries M(μ), N(μ) to produce the renormalized tensors. The reshaping makes the bottleneck contraction scale as O($r^{2}$ $χ^{7}$) instead of O($χ^{9}$), and when r = χ the method reduces exactly to the ATRG. Numerically, the claim is that at r = 7 the free energy and critical temperature of the four-dimensional Ising model agree with ATRG results to better than 0.0013% and 0.1% respectively at χ = 54, with GPU timing scaling better than O($χ^{6}$).

Load-bearing premise

The argument assumes that one fixed oversampling parameter, r = 7, keeps the truncated triad network accurate enough over the whole range of bond dimensions and for the observables studied; this is only checked empirically for the four-dimensional Ising free energy and transition temperature.

Editorial extensions

If this is right

The same reduction should make bond dimensions well beyond χ ≈ 50 accessible for four-dimensional tensor network studies on existing hardware.
Larger r buys accuracy at a predictable multiplicative cost, and r = χ recovers the ATRG exactly, giving a continuous interpolation between cheap approximation and exact ATRG.
The GPU implementation of Triad-ATRG scales better than O(χ^6), so the method is particularly attractive for GPU-accelerated lattice computations.
Because the accuracy loss at r = 7 is far smaller than the bond-dimension truncation error, Triad-ATRG can replace ATRG as the workhorse in four-dimensional studies where only the final converged value matters.

Reading between the lines

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

If the cost scaling holds in other models, the r parameter becomes a user-tunable dial that could be adjusted per RG step or per observable, potentially improving accuracy beyond the fixed r = 7 used here.
The method's reliance on a single oversampling factor suggests a natural stress test: apply Triad-ATRG to models with long-range couplings or fermionic sign structure, where the bond-entanglement structure differs from the Ising model.
The improved GPU scaling hints that the remaining bottleneck is memory bandwidth rather than arithmetic, so mixed-precision or blocked tensor contractions might extend the reach further.
Since Triad-ATRG cannot exceed ATRG accuracy even as r grows, any future four-dimensional result that needs beyond-ATRG accuracy would still require a different improvement, such as better squeezers or bond-swapping alternatives.

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.

Desk Editor's Note

read the letter

Punchline: this is a solid algorithmic extension, not a breakthrough. The authors graft the triad/MDTRG representation onto the four-dimensional ATRG unit-cell tensor and get a real cost reduction in the contraction step from O(χ^9) to O(r^2 χ^7). With r=7 and χ up to about 54 the savings are practical, and the 4D Ising free energy tracks ATRG well enough for a first demonstration. The paper is also admirably honest about where the method stops being exact: it says plainly that r=χ reduces to ATRG and that Triad-ATRG cannot beat ATRG in accuracy.

What's new: previous triad/MDTRG work stayed in three dimensions, and ATRG had not been given the triad treatment. The cost table is internally consistent (bond swapping O(q r χ^6), triad construction O(χ^7), squeezer O(min(χ^7, r^2 χ^6)), contraction O(r^2 χ^7)). The numerics are reproducible in outline, and r and q are algorithmic truncation parameters rather than fitted knobs, so there is no circularity problem.

The soft spots, in rough order of importance. First, the Section 3 scaling text has the exponents exactly backwards: it says ATRG is O(χ^7) and Triad-ATRG is O(χ^9), while Figure 5 and Table 1 say the opposite. That is a typo, but it needs to be fixed. Second, the GPU claim is not supported: the paper says Triad-ATRG scales smaller than O(χ^6) on two GPUs, which is inconsistent with the asymptotic O(r^2 χ^7) cost. The measured range is too narrow and there are no error bars; this needs a much more careful statement. Third, the accuracy evidence is weaker than the abstract promises. Table 2 shows the gap between Triad-ATRG and ATRG is about 4e-5 at χ=38 and 6e-5 at χ=54, with no trend toward zero. If the triad truncation error were controlled by the growing 7χ bond, the deviation should shrink with χ; it does not. So the right claim is 'agreement to about 0.001% in the range we tested', not 'maintained convergence accuracy'. Adding one comparison to a more accurate reference (HOTRG or a Monte Carlo value) at least at a couple of χ values would settle whether the offset is a finite-χ artifact or a real limit. I don't think this is fatal: the cost reduction is the deliverable, and the free energy agreement is good enough for a proceedings-style paper.

Who it's for: tensor network practitioners working on 4D systems, especially finite-density QCD, who want a cheaper ATRG variant. I'd send it to peer review; a competent referee can sort out the text and ask for the extra benchmark. The method deserves to be in the literature.

Referee Report

5 major / 7 minor

Summary. The paper proposes Triad-ATRG, a variant of the four-dimensional anisotropic tensor renormalization group (ATRG) in which the unit-cell tensor is decomposed into a triad network using oversampled isometries and randomized SVD, following the ideas of MDTRG/Triad-MDTRG. The claimed advantage is a reduction of the contraction-step cost from O(chi^9) in ATRG to O(r^2 chi^7) in Triad-ATRG, where r is an oversampling parameter. The authors benchmark the method on the four-dimensional Ising model, comparing free energies and the phase transition temperature T_c with ATRG results for r=7 and 10, and they report GPU timing measurements that suggest further practical speedups. The manuscript concludes that Triad-ATRG reproduces ATRG accuracy for the free energy and T_c while substantially reducing computational cost.

Significance. If the central claim holds, Triad-ATRG would be a practically useful step toward larger bond dimensions in four-dimensional tensor network calculations, an important direction for finite-density QCD and other four-dimensional systems. The paper is honest in stating that r=chi reduces to ATRG and that Triad-ATRG cannot exceed ATRG accuracy, which correctly frames the method as a cheaper approximation of ATRG. The numerical implementation is reproducible in structure and the cost tables are self-consistent. The most valuable contribution would be the demonstrated numerical scaling on CPU/GPU for chi up to 54, if the accuracy gap to ATRG is indeed under control; however, as detailed below, the evidence for maintained convergence accuracy is weaker than the text claims, and the scaling statements in Section 3 are internally inconsistent and need correction.

major comments (5)

[Sec. 3, text after Fig. 5 and Fig. 6] The scaling statements in Section 3 are reversed and inconsistent. The paragraph after Fig. 5 says 'the ATRG scales as O(chi^7), while the Triad-ATRG scales as O(chi^9)', which contradicts both Table 1 and the central claim of the paper. The figure caption for Fig. 5 also labels the lines as chi^9 and chi^7, which matches the reversal. These statements must be corrected so that ATRG is O(chi^9) and Triad-ATRG is O(r^2 chi^7) in the CPU scaling plot. This is a load-bearing presentation error because the paper's main selling point is the reduced asymptotic cost of Triad-ATRG.
[Sec. 3, Fig. 6 caption and surrounding text] The GPU scaling claim 'Triad-ATRG scales as smaller than O(chi^6)' is not supported by the algorithm's asymptotic cost. The theoretical cost of the contraction step remains O(r^2 chi^7), and the squeezer step is at best O(min(chi^7, r^2 chi^6)); no GPU-specific asymptotic reduction is derived in the paper. The empirical timing slope on two GPUs over a limited chi range may reflect GPU memory access patterns or small-r prefactors, but the text overinterprets it as an asymptotic scaling result. Please rephrase this as an empirical observation and clarify that the algorithmic cost remains O(r^2 chi^7).
[Sec. 3, Table 2 and Fig. 4] The accuracy evidence for 'maintaining the convergence accuracy' is not demonstrated by the data. The absolute difference between ATRG and Triad-ATRG (r=7) free energies is 4.41e-5 at chi=38 and 6.43e-5 at chi=54, with values fluctuating between 2.4e-5 and 7.3e-5 and no decreasing trend. If the triad truncation error were controlled by the growing effective bond dimension (r chi), one would expect the gap to shrink with chi; instead the data suggest a roughly constant offset. Since the paper correctly notes that Triad-ATRG cannot be more accurate than ATRG, the target is reproducing ATRG, and a non-shrinking gap means the approximation is not shown to be systematically improvable in chi alone at fixed r. The paper should either present the gap as a function of chi and discuss its trend, or compare both methods against an independent reference such as higher-precision HOTRG or a known exact/dual result, to establish what the constant offset means.
[Sec. 4, Summary] The summary statement that 'the approximation adopted by the new method to reduce costs does not affect the accuracy in the evaluation of the free energy and phase transition point' is too strong. Table 2 shows a systematically positive difference between ATRG and Triad-ATRG free energies (ATRG lower in every row), and Fig. 7 shows that not all T_c values converge with chi. The correct claim is that the deviations from ATRG are small in the tested range (0.0013% for free energy, under 0.1% for T_c), not that accuracy is unaffected. This wording should be weakened to match the actual evidence.
[Sec. 2, Eqs. (4)-(11) and squeezer definition] The derivation of the triad representation and the squeezers is sketched rather than fully specified. In particular, Eq. (12) defines squeezers M(mu), N(mu) by a cost function, but the actual construction of M and N from the full triad tensors is not described; the text says 'all fundamental tensors must be included' but does not give the SVD or QR steps. Since the claim of reduced cost depends on the way the squeezers are built, this should be documented at least at the level of a pseudo-code or a reference to a specific equation in Ref. [13]. Without this, a reader cannot verify the O(min(chi^7, r^2 chi^6)) entry in Table 1.

minor comments (7)

[Abstract] The abstract says 'maintaining the convergence accuracy of the free energy', which is not exactly what is shown; the data show small deviations from ATRG, not convergence of the free energy to a reference value. Please adjust the wording.
[Sec. 1, Introduction] The sentence 'both the MDTRG and the Triad-MDTRG have successfully achieved free energy calculations consistent with the previous study done with the HOTRG in the three-dimensional Ising model' states the MDTRG result, but the paper does not apply MDTRG to four dimensions; this is fine, but the text could be clearer that the four-dimensional application is new.
[Sec. 3, Fig. 6] In Fig. 6, the label 'smaller than O(chi^6)' is ambiguous; specify whether the fit gives an effective exponent and what chi range was used for the fit.
[Sec. 3, Fig. 7] The error bars in Fig. 7 are described as resulting from temperature resolution, but the text does not state the temperature step size or how T_c is interpolated. This should be stated so the 0.1% claim can be assessed.
[Sec. 3, text after Fig. 4] The sentence 'the difference between the Triad-ATRG and ATRG. was only 0.0013%' contains a stray period after 'ATRG' and should be corrected.
[Sec. 2, Eq. (1)] The notation for the bond indices i_mu(n) and j_mu(n) is introduced but not defined consistently in Eq. (1); a short explanation of the index conventions would help readability.
[Sec. 2, text after Eq. (13)] The phrase 'we use the notation /gamma, i_1, k in the sense of summing except for gamma, i_1, k' is confusing; please spell out the summation indices explicitly in Eq. (13) or add a defining sentence.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: Triad-ATRG is a genuinely simplified approximation of ATRG, benchmarked honestly against ATRG without fitted parameters.

full rationale

The paper constructs a triad representation of ATRG by inserting oversampled isometries derived from SVD of the canonical unit-cell tensor (Eqs. (2)-(11)). The computational cost claim O(r^2 chi^7) follows from explicit tensor-contraction counting and is not a restatement of an input. The accuracy claim is that Triad-ATRG with r=7 reproduces ATRG's free energy and Tc; this is a benchmarking choice, not circular reasoning, because the algorithm is not fitted to ATRG outputs and r is an algorithmic oversampling hyperparameter, not a fitted parameter. The paper explicitly notes that r=chi reduces to ATRG, so the approximation cannot exceed ATRG accuracy; this honest statement removes any pretense that ATRG accuracy is derived independently. The method inherits its decomposition and RSVD components from prior work [10,11], but those are external citations by other authors and are not self-citations. The absence of a shrinking gap between Triad-ATRG and ATRG in Table 2 is a correctness/evidence concern, not a circularity concern. Overall, the derivation chain is self-contained and no step reduces to its own input by construction.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The method's cost and accuracy claims rest on standard tensor-network truncation theory plus the hand-chosen parameter r=7 and the unspecified RSVD iteration count q. No new physical entities are introduced. Empirical validation is limited to the four-dimensional Ising model, with accuracy benchmarked only against ATRG.

free parameters (2)

Oversampling parameter r = 7 (also 10 for phase transition point)
Controls the trade-off between cost O(r^2 χ^7) and accuracy; chosen by hand, not fitted to physical data, but central to the claimed speedup.
RSVD iteration count q = unspecified
Appears in the cost formula O(q r χ^6) for bond swapping and in the randomized SVD; the paper never reports its value, so cost and reproducibility are underdetermined.

assumptions (5)

domain assumption The ATRG unit-cell tensor Γ after bond swapping is in canonical form A X σ Y D, so the optimal isometries U_A are obtained from the SVD of A X σ (Eq. 3).
Invoked in Section 2 to avoid a full SVD of Γ; relies on Ref. [12].
domain assumption Inserting oversampled isometries and truncating to rχ singular values preserves the partition function to sufficient accuracy.
Standard TRG truncation assumption; validated only empirically for the four-dimensional Ising model in Section 3.
standard math The four-dimensional Ising partition function is exactly represented by the initial tensor in Eqs. (15)-(17).
Standard exact tensor-network mapping of the Ising model.
domain assumption Randomized SVD with q QR iterations yields the dominant singular subspaces accurately enough for the claimed cost-accuracy trade-off.
Basis for the O(q r χ^6) bond-swapping cost in Table 1; q is never specified.
domain assumption The squeezers M and N from Ref. [13] can be computed and applied with the stated costs and preserve accuracy.
Used in Section 2; no derivation of the O(min(χ^7, r^2 χ^6)) cost is given.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Applying the Triad network representation to four-dimensional ATRG method." pith.science (2026). https://pith.science/paper/Y4YQXF73

@misc{pith2026241214104,
  author       = {Pith},
  title        = {Pith review of: Applying the Triad network representation to four-dimensional ATRG method},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y4YQXF73}},
  note         = {Machine review of arXiv:2412.14104}
}

read the original abstract

Anisotropic Tensor Renormalization Group (ATRG) is a powerful algorithm for four-dimensional tensor network calculations. However, the larger bond dimensions are known to be difficult to achieve in practice due to the higher computational cost. Adopting the methods of the minimally decomposed TRG and its triad prescriptions, we construct a triad representation of the four-dimensional ATRG by decomposing the unit-cell tensor. We observe that this combining approach can significantly improve the computational cost even with maintaining the convergence accuracy of the free energy in the four-dimensional Ising model. In addition, we also show that a further improvement can be achieved in terms of the computational cost when our proposed approach is implemented in parallel on GPUs.

Figures

Figures reproduced from arXiv: 2412.14104 by the authors.

**Figure 2.** Schematic view of the calculation process for the contraction part in the Triad-ATRG. The bottleneck part has a cost of 𝑂(𝑟 2𝜒 7 ) [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Schematic overview of the Triad-ATRG algorithm. should be noted that if 𝑟 is taken large, the contraction step may become a bottleneck. It is worth pointing out that when 𝑟 = 𝜒, the Triad-ATRG algorithm corresponds to the ATRG, implying that it cannot achieve higher accuracy than the ATRG [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: The approximated free energy of four-dimensional Ising model by the ATRG and Triad-ATRG with 𝑟 = 7. s where 𝑇 is the initial tensor given by 𝑇𝑖1 (𝑛)𝑖2 (𝑛)𝑖3 (𝑛)𝑖4 (𝑛) 𝑗1 (𝑛) 𝑗2 (𝑛) 𝑗3 (𝑛) 𝑗4 (𝑛) = ∑︁ 2 𝑎=1 𝑊𝑎𝑖1 (𝑛)𝑊𝑎𝑖2 (𝑛)𝑊𝑎𝑖3 (𝑛)𝑊𝑎𝑖4 (𝑛)𝑊𝑎 𝑗1 (𝑛)𝑊𝑎 𝑗2 (𝑛)𝑊𝑎 𝑗3 (𝑛)𝑊𝑎 𝑗…

**Figure 6.** Figure 6: Scalings of computational time on two GPUs by the ATRG and Triad-ATRG with 𝑟 = 7. GPU parallelization algorithm [14] can be applied to both methods. In this study, two GPUs are used, and the resulting scaling of the computational time is shown in [PITH_FULL_IMAGE:figu…

**Figure 7.** Figure 7: The phase transition point of the four-dimensional Ising model by the the ATRG and Triad-ATRG. degeneracy of the ground state as discussed in Ref. [15]. The definition of 𝑋 is given by 𝑋 (𝑚) = (Tr𝐴 (𝑚) ) 2 Tr(𝐴(𝑚) ) 2 with 𝐴 (𝑚) 𝑘𝑙 = ∑︁ 𝑖1,𝑖2,𝑖3 𝑇 (𝑚) 𝑖1𝑖2𝑖3 𝑘𝑖1𝑖2𝑖3𝑙 ,…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

15 extracted references · 2 canonical work pages

[13]

S. Iino, S. Morita and N. Kawashima,Boundary Tensor Renormalization Group,Phys. Rev. B 100 (2019) 035449 [1905.02351]

arXiv 2019
[1]

Levin and C.P

M. Levin and C.P. Nave,Tensor renormalization group approach to 2D classical lattice models,Phys. Rev. Lett.99 (2007) 120601 [cond-mat/0611687]

arXiv 2007
[2]

Z.Y. Xie, J. Chen, M.P. Qin, J.W. Zhu, L.P. Yang and T. Xiang,Coarse-graining renormalization by higher-order singular value decomposition, Phys. Rev. B86 (2012) 045139 [1201.1144]

arXiv 2012
[3]

Akiyama, Y

S. Akiyama, Y. Kuramashi, T. Yamashita and Y. Yoshimura,Phase transition of four-dimensional Ising model with higher-order tensor renormalization group,Phys. Rev. D 100 (2019) 054510 [1906.06060]

arXiv 2019
[4]

Phase transition of four-dimensional Ising model with tensor network scheme

S. Akiyama, Y. Kuramashi, T. Yamashita and Y. Yoshimura,Phase transition of four-dimensional Ising model with tensor network scheme, PoS LATTICE2019(2019) 138 [1911.12954]

work page Pith review arXiv 2019
[5]

Adachi, T

D. Adachi, T. Okubo and S. Todo,Anisotropic Tensor Renormalization Group,Phys. Rev. B 102 (2020) 054432 [1906.02007]

arXiv 2020
[6]

Akiyama, D

S. Akiyama, D. Kadoh, Y. Kuramashi, T. Yamashita and Y. Yoshimura,Tensor renormalization group approach to four-dimensional complex𝜙4 theory at finite density, JHEP 09 (2020) 177 [2005.04645]

arXiv 2020
[7]

Akiyama, Y

S. Akiyama, Y. Kuramashi, T. Yamashita and Y. Yoshimura,Restoration of chiral symmetry in cold and dense Nambu–Jona-Lasinio model with tensor renormalization group,JHEP 01 (2021) 121 [2009.11583]

arXiv 2021

Show all 15 references

[8]

Akiyama and Y

S. Akiyama and Y. Kuramashi,Critical endpoint of (3+1)-dimensional finite densityZ3 gauge-Higgs model with tensor renormalization group,JHEP 10 (2023) 077 [2304.07934]

2023 arXiv
[9]

Kadoh and K

D. Kadoh and K. Nakayama,Renormalization group on a triad network, 1912.02414

1912 arXiv
[10]

Nakayama,Application of the projective truncation and randomized singular value decomposition to a higher dimension, PoS LATTICE2023(2024) 029 [2401.06389]

K. Nakayama,Application of the projective truncation and randomized singular value decomposition to a higher dimension, PoS LATTICE2023(2024) 029 [2401.06389]

2024 arXiv
[11]

Nakayama,Randomized higher-order tensor renormalization group, 2307.14191

K. Nakayama,Randomized higher-order tensor renormalization group, 2307.14191

arXiv
[12]

Akiyama, R.G

S. Akiyama, R.G. Jha and J. Unmuth-Yockey,SU(2) principal chiral model with tensor renormalization group on a cubic lattice, Phys. Rev. D110 (2024) 034519 [2406.10081]

2024 arXiv
[14]

Jha and A

R.G. Jha and A. Samlodia,GPU-Acceleration of Tensor Renormalization with PyTorch using CUDA, Comput. Phys. Commun.294 (2024) 108941 [2306.00358]

2024 arXiv
[15]

Gu and X.-G

Z.-C. Gu and X.-G. Wen,Tensor-Entanglement-Filtering Renormalization Approach and Symmetry Protected Topological Order,Phys. Rev. B80(2009) 155131 [0903.1069]. 9

2009 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [13]

S. Iino, S. Morita and N. Kawashima,Boundary Tensor Renormalization Group,Phys. Rev. B 100 (2019) 035449 [1905.02351]

arXiv 2019

[2] [1]

Levin and C.P

M. Levin and C.P. Nave,Tensor renormalization group approach to 2D classical lattice models,Phys. Rev. Lett.99 (2007) 120601 [cond-mat/0611687]

arXiv 2007

[3] [2]

Z.Y. Xie, J. Chen, M.P. Qin, J.W. Zhu, L.P. Yang and T. Xiang,Coarse-graining renormalization by higher-order singular value decomposition, Phys. Rev. B86 (2012) 045139 [1201.1144]

arXiv 2012

[4] [3]

Akiyama, Y

S. Akiyama, Y. Kuramashi, T. Yamashita and Y. Yoshimura,Phase transition of four-dimensional Ising model with higher-order tensor renormalization group,Phys. Rev. D 100 (2019) 054510 [1906.06060]

arXiv 2019

[5] [4]

Phase transition of four-dimensional Ising model with tensor network scheme

S. Akiyama, Y. Kuramashi, T. Yamashita and Y. Yoshimura,Phase transition of four-dimensional Ising model with tensor network scheme, PoS LATTICE2019(2019) 138 [1911.12954]

work page Pith review arXiv 2019

[6] [5]

Adachi, T

D. Adachi, T. Okubo and S. Todo,Anisotropic Tensor Renormalization Group,Phys. Rev. B 102 (2020) 054432 [1906.02007]

arXiv 2020

[7] [6]

Akiyama, D

S. Akiyama, D. Kadoh, Y. Kuramashi, T. Yamashita and Y. Yoshimura,Tensor renormalization group approach to four-dimensional complex𝜙4 theory at finite density, JHEP 09 (2020) 177 [2005.04645]

arXiv 2020

[8] [7]

Akiyama, Y

S. Akiyama, Y. Kuramashi, T. Yamashita and Y. Yoshimura,Restoration of chiral symmetry in cold and dense Nambu–Jona-Lasinio model with tensor renormalization group,JHEP 01 (2021) 121 [2009.11583]

arXiv 2021