Neural and Tensor-Network Representations of Z N Cluster and Dipolar-cluster SPT States
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Condensed Matter. The day in condensed matter: materials, phases and the measurements under them.
Kai: We're looking at the paper: "Neural and Tensor-Network Representations of Z N Cluster and Dipolar-cluster SPT States".
Mira: Here's the gist of the paper. The ZN cluster-state wavefunction, a paradigmatic example of symmetry-protected topological (SPT) order with ZN ×ZN symmetry, is expressed in various equivalent ways <ref:2604.06741#pg2> The P-representation is identified as the efficient way to express cluster and dipolar cluster state’s wavefunctions <ref:2604.06741#pg4> Employing the restricted Boltzmann machine (RBM) scheme to re-write the interaction matrix in the P-representation in terms of neural weight matrices allows for the development of neural quantum state (NQS) and matrix product state (MPS) representations <ref:2604.06741#pg4> The tensor product state (TPS) representation is derived for the dipolar cluster state where each local tensor carries three virtual indices connecting a given site to two nearest neighbors and one further neighbor <ref:2604.06741#pg4>.
Kai: Representational Framework The P-representation involves the product of a local projector P s and the interaction matrix omega, where omega can be decomposed as a product of two weight matrices, e.g., omega = WWt <ref:2604.06741#pg3>.
Mira: This leads to the NQS representation where the wavefunction is expressed as Ψ(s) = Tr
P s1WWtP s2omega∗P s3omega···P sLomega∗: <ref:2604.06741#pg6>.
Kai: The MPS representation is obtained by setting A s2 j−1 = omega∗P s2 j−1 and A s2 j = omegaP s2 j <ref:2604.06741#pg4>.
Title and authors: Kai: Okay, shifting focus slightly, the paper "Neural and Tensor-Network Representations of Z N Cluster and Dipolar-cluster SPT States" brings in these authors: Seungho Lee, Daesik Kim, Hyun-Yong Lee, Jung Hoon Han.
Mira: They're clearly coming from a background that connects machine learning with condensed matter theory because the work is bridging those two worlds.
Lev: I’m curious about the specific expertise they bring to this problem; does this team have experience mapping these kinds of symmetries onto actual error-correction codes?
Kai: Lev, you're right, that's a big part of it. They are dealing with Z N cluster states which are textbook examples of symmetry-protected topological order, and they are looking at how to represent them using P-representation.
Mira: What I find interesting is the focus on generalizing results from simpler systems like Z2 cluster states up to arbitrary Z N variables, which suggests a very robust framework.
Lev: Generalizing to arbitrary Z N is tough when you're talking about real hardware; we usually have to stick to small groups for error correction codes, so that generalization is key.
Kai: So the core idea here isn't just describing a known state, but developing a universal representation method using neural networks that works across different symmetry groups.
Mira: Exactly, and they are also looking at how this P-representation helps analyze symmetry invariance through those push-through conditions.
Lev: Analyzing fractionalization through those mathematical conditions is theoretical work; it's helpful to see if those symmetries actually map onto observable quantities in a physical setup.
The paper's summary: Kai: So, summarizing what the authors are showing in "Neural and Tensor-Network Representations of Z N Cluster and Dipolar-cluster SPT States," they take this Z N cluster state wavefunction, which has Z N by Z N symmetry, and express it in several equivalent forms.
Mira: They identify the P-representation as being the most efficient way to write down these cluster and dipolar cluster state wavefunctions.
Lev: Efficiency in representation is one thing; on real hardware, efficiency also translates directly into required coherence times and gate fidelity, so that's a practical concern for us.
Kai: The paper shows how using the restricted Boltzmann machine scheme to rewrite the interaction matrix in that P-representation allows them to get both neural quantum state and matrix product state representations of the same state.
Mira: They also derive a tensor product state representation specifically for the dipolar cluster state, where each local tensor has three virtual indices connecting a site to two nearest neighbors and one further neighbor.
Lev: That three-index structure is getting complicated quickly; it makes me wonder about the scaling of the required bond dimension in the MPS representation when you use that kind of tensor.
Kai: The paper also proposes an interpretation for the KramersWannier operator as being like the dipolar version of the discrete Fourier transform on Z N variables, which gives intuition about why that map isn't invertible.
Mira: That intuition about non-invertibility is useful because it connects a mathematical constraint on a variable, specifically dj equal to sj minus sj plus one, to physical limitations in the state space.
The paper's improvements: Kai: Now let's talk about what they suggest as improvements in this work. They focus on showing how the P-representation helps with symmetry analysis by performing manipulations in that representation, which they call the Prepresentation.
Mira: This is key because it shows that you can analyze symmetry invariance through these manipulations just as easily, if not more so, using this P-representation compared to traditional methods.
Lev: If the analysis is easier on paper, how much does that translate to reducing the complexity of implementing error correction gates in a physical setup?
Kai: The paper also shows how the NQS leads directly to an MPS by reorganizing weight functions around a site instead of just around a bond. That's a structural improvement in how you think about the entanglement.
Mira: And they show that this framework is applicable to other mSPTs, like those protected by quadrupolar and exponential symmetries, which expands the scope beyond just the Z N cluster state.
Lev: Expanding the scope is fine theoretically, but we need to know if those new symmetry types have corresponding efficient tensor network representations that are actually viable for simulation.
Kai: Furthermore, they provide closed-form expressions for the NQS weights in all three cases—NQS, MPS, and TPS—which is a nice simplification for calculation.
Conclusion: Kai: So to wrap up on "Neural and Tensor-Network Representations of Z N Cluster and Dipolar-cluster SPT States," the main point is that they've shown how this P-representation allows us to build neural quantum states, which then lead to MPS, and in some cases, a TPS representation.
Mira: And they've proven that this framework is useful for analyzing symmetry fractionalization through those push-through conditions of the P-tensor.
Lev: From my side, I think the real utility for hardware is seeing how the KramersWannier duality operator can be written as a matrix product operator, which helps us construct MPO representations for operators we might want to run on NISQ devices.
Kai: And they've also shown that they can generalize earlier Z2 results to arbitrary Z N variables and show that certain dipolar SPT states naturally require the TPS representation as a more efficient way to represent them via DMRG calculations.
Mira: So, the paper provides a systematic path for analyzing symmetry invariance and fractionalization through these P-tensors, giving us closed-form expressions for the NQS weights in all three cases.
Lev: The limitation they mention is that while the TPS representation is more efficient representatively, its exact contraction can be computationally more expensive than an MPS representation.
Kai: That’s a fair caveat; it's a trade-off between the structural clarity of the tensor network and raw computational speed for large systems.
Mira: But overall, this work gives us tools to systematically tackle these types of mSPTs, which is useful for understanding how these complex symmetries behave in physical matter.
Lev: We'll keep an eye on how this framework translates into practical error correction schemes as the authors move forward with applying it to real hardware.
Department of Physics, Sungkyunkwan University · Institute of Basic Science, Sungkyunkwan University · Division of Semiconductor Physics, Korea University · Department of Applied Physics, Graduate School, Korea University
cond-mat.dis-nn, cond-mat.str-el, physics.chem-ph
Submitted: 2026-04-08
Updated: 2026-11-12
Comments: 21 pages, 10 figures
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Importance score: 84/100
The gist: The ZN cluster-state wavefunction, a paradigmatic example of symmetry-protected topological (SPT) order with ZN ×ZN symmetry…
Key concepts
- P-representation
- This is an efficient way to express cluster and dipolar cluster state wavefunctions by involving a local projector P s and the interaction matrix Ω. It serves as a unifying bridge connecting different quantum representations like NQS, MPS, and TPS.
- NQS/MPS/TPS Representations
- These are different mathematical ways to describe the quantum state. NQS (Neural Quantum State) is derived from the P-representation, which can be rewritten as an MPS by relating weight matrices. TPS (Tensor Product State) is a specific representation for dipolar cluster states where local tensors have three virtual indices.
- Kramers-Wannier Duality
- The ZN Kramers-Wannier operator acts as a noninvertible symmetry, interpreted as the dipolar generalization of the discrete Fourier transform on ZN variables. This explains why the map is noninvertible because a constraint exists on the target variable dj.
- Symmetry Analysis via Push-through Conditions
- The paper shows how to check if cluster state wavefunctions respect global symmetries (like C1 and C2) by using specific 'push-through conditions' applied to the P-representation, simplifying the analysis of symmetry invariance.
Terminology
Summary
The ZN cluster-state wavefunction, a paradigmatic example of symmetry-protected topological (SPT) order with ZN ×ZN symmetry, is expressed in various equivalent ways <ref:2604. p. 006741#pg2> The P-representation is identified as the efficient way to express cluster and dipolar cluster state’s wavefunctions <ref:2604. p. 006741#pg4> Employing the restricted Boltzmann machine (RBM) scheme to re-write the interaction matrix in the P-representation in terms of neural weight matrices allows for the development of neural quantum state (NQS) and matrix product state (MPS) representations <ref:2604. p. 006741#pg4> The tensor product state (TPS) representation is derived for the dipolar cluster state where each local tensor carries three virtual indices connecting a given site to two nearest neighbors and one further neighbor <ref:2604. p. 006741#pg4>.
Representational Framework
The P-representation involves the product of a local projector P s and the interaction matrix omega, where omega can be decomposed as a product of two weight matrices, e.g., omega = WWt <ref:2604. p. 006741#pg3>. This leads to the NQS representation where the wavefunction is expressed as Ψ(s) = Tr[P s1WWtP s2omega∗P s3omega···P sLomega∗] <ref:2604. p. 206741#pg6>. The MPS representation is obtained by setting A s2 j−1 = omega∗P s2 j−1 and A s2 j = omegaP s2 j <ref:2604. p. 706741#pg4>. The P-representation serves as a unifying bridge between NQS and tensor-network-type representations <ref:2604. p. 006741#pg4>.
Symmetry Analysis
The invariance of the cluster state wavefunction under global symmetries C1 and C2 can be checked in the Prepresentation, showing that the transformation results in a form equivalent to the original one through push-through conditions <ref:2604.06741#pg4>. For Type-I dipolar cluster states, invariance under C and D is confirmed by showing that the modified projector P s obeys the push-through condition P s+1 = Z†XPsZ†X <ref:2604. p. 406741#pg4>. The proof for D symmetry in Type-I dCSII involves a graphical argument demonstrating how terms cancel out except for one factor of XZ† on the right side of each omega <ref:2604. p. 406741#pg4>.
Tensor Network Efficiency
For the Type-II dipolar cluster state (dCSII), the P-representation naturally yields a TPS representation with local tensors carrying three virtual indices, which is more compact at the level of local tensors than the MPS representation <ref:2604.06741#pg4>. Although exact contraction of the TPS is computationally more expensive than MPS, its advantage is primarily representational as it reveals the natural tensor structure dictated by modulated symmetries <ref:2604. p. 706741#pg4>.
Kramers-Wannier Duality
The ZN Kramers-Wannier (KW) operator implements a noninvertible symmetry and can be interpreted as the dipolar generalization of the discrete Fourier transform on ZN variables <ref:2604.06741#pg4>. This interpretation explains why the KW map is non-invertible because the target variable dj ≡ sj − sj+1 has a constraint, which covers only 1/N-th of the target space <ref:2604. p. 806741#pg4>. The MPO representation of the KW operator can be found in terms of P2-representation, illustrating its utility for constructing NIS operators as an MPO <ref:2604. p. 906741#pg4>.
Conclusion and Generalization
The work demonstrates that the NQS leads to MPS by re-organizing weight functions around a site rather than around a bond <ref:2604.06741#pg4>. The P-representation generalizes easily to projectors having multiple virtual legs, resulting in the TPS representation of the wavefunction <ref:2604. p. 606741#pg4>. This framework is applicable to other mSPTs such as those protected by quadrupolar and exponential symmetries <ref:2604. p. 606741#pg4>. The P-representation facilitates the analysis of symmetry invariance and its fractionalization through push-through conditions of the P-tensor <ref:2604. p. 606741#pg4>. The NQS construction provides closed-form expressions for the NQS weights in all three cases <ref:2604. p. 906741#pg4>. The explicit MPO form of the Kramers–Wannier duality operator generalizes earlier Z2 results to arbitrary ZN <ref:2604. p. 906741#pg4>. Perturbations of the cluster Hamiltonians can be viewed as renormalizations of the interaction matrix omega, or the weight matrix W <ref:2604. p. 906741#pg4>. The nonlinearity structure of the weight functions is tied to the entanglement content of the state <ref:2604. p. 706741#pg4>. Exploring these directions remains an interesting avenue for future research <ref:2604. p. 006741#pg4>. The paper also shows that certain dipolar SPT states naturally require TPS as the more efficient way to represent its many-body state via DMRG calculations <ref:2604. p. 106741#pg4>. The NQS/TPS construction for dCSII is primarily a representational advantage rather than computational efficiency <ref:2604. p. 706741#pg4>.
Appendix Details
The derivation of the weight matrix W(s,h) for the ZN cluster state involves finding an integer r satisfying specific conditions to obtain the RBM representation <ref:2604.06741#pg3>. The proof of symmetry invariance for the MPS representation in Appendix C is more involved than in the NQS representation <ref:2604. p. 206741#pg4>. The proof of symmetry for dCSII under D1 involves summing over hidden variables to recover the original wavefunction <ref:2604. p. 506741#pg4>. The transformation of the P-representation under C†2 is shown through a sequence of index relabelings and property applications <ref:2604. p. 4 p. 406741#pg4>. The proof for D†2 involves showing how the transformation leads to a specific phase factor that restores the original wavefunction <ref:2604. p. 306741#pg4>.
Appendix E Details
The local MPO representation of Kc is derived by combining two traces, which can be re-written as a single trace involving products of projectors and weight matrices <ref:2604.06741#pg4>. This MPO representation is obtained by examining how the MPS representation transforms under the KW operator <ref:2604. p. 006741#pg4>. The local tensor (K j c) is defined as a product of terms involving W and W† matrices <ref:2604. p. 1 p. 106741#pg4>. The MPO representation of Kc can also be derived by starting from the MPO representation of the ZN KW operator in Eq. p. 0 (4.3) <ref:2604. p. 006741#pg4>.
Appendix F Details
The type-II dipolar cluster model admits a noninvertible symmetry Kd which is shown to be equivalent to a quadrupole Fourier transformation <ref:2604.06741#pg4>. The Kd-invariance of the dCSII ground state immediately implies the first four fusion rules satisfied by the symmetry generators <ref:2604. p. 8 p. 806741#pg4>. The MPO representation of Kd is given in terms of TPS representations, where each bond index is a collection of four ZN numbers <ref:2604. p. 9 p. 906741#pg4>. The Kd-invariance proof for dCSII shows that the image lies entirely in the sector C1 = C2 = D1 = D2 = 1 <ref:2604. p. 6 p. 606741#pg4>. The Kd-invariance of the dCSII wavefunction immediately implies the first four of the following fusion rules satisfied by the symmetry generators <ref:2604. p. 806741#pg4>.
References
[1] G. Carleo and M. Troyer, Solving the Quantum Many-Body Problem with Artificial Neural Networks, Science 355, 602 (2017) <ref:2604. p. 906741#pg4>
[2] L. L. Viteritti, R. Rende, and F. Becca, Transformer Variational Wave Functions for Frustrated Quantum Spin Systems, Phys. p. 0 Rev. Lett. 130, 236401 (2023) <ref:2604. p. 006741#pg4>
[3] D.-L. Deng, X. Li, and S. Das Sarma, Machine learning topological states, Phys. p. 0 Rev. B 96, 195145 (2017) <ref:2604. p. 006741#pg4>
[4] D.-L. Deng, X. Li, and S. Das Sarma, Quantum Entanglement in Neural Network States, Phys. p. 0 Rev. X 7, 021021 (2017), arXiv:1701. p. 004844 <ref:2604. p. 006741#pg4>
[5] X. Gao and L.-M. Duan, Efficient representation of quantum many-body states with deep neural networks, Nature Communications 8, 662 (2017) <ref:2604. p. 906741#pg4>
[6] I. Glasser, N. Pancotti, M. August, I. D. Rodr´ıguez, and I. Cirac, Neural-network quantum states, string-bond states, and chiral topological states, Phys. p. 0 Rev. X 8, 011006 (2018) <ref:2604. p. 006741#pg4>
[7] S. R. Clark, Unifying neural-network quantum states and correlator product states via tensor networks, J. p. 0 Phys. A: Math. Theor. 51, 135301 (2018) <ref:2604. p. 006741#pg4>
[8] J. Chen, S. Cheng, H.-Y Xie, L Wang and T Xiang, Equivalence of restricted Boltzmann machines and tensor network states, Phys. p. 0 Rev. B 97, 085104 (2018) <ref:2604. p. 006741#pg4>
[9] Z.-A Jia, Y.-H Zhang, Y.-C Wu, L Kong, G-C Guo and G-P Guo, Efficient machine-learning representations of a surface code with boundaries, defects, domain walls, and twists, Phys. p. 0 Rev. A 99, 012307 (2018) <ref:2604. p. 006741#pg4>
[10] Z.-A Jia, L Wei, Y-C Wu, G-C Guo and G.-P Guo, Entanglement area law for shallow and deep quantum neural network states, New Journal of Physics 22, 053022 (2020) <ref:2604. p. 006741#pg4>
[11] O. Sharir, A. Shashua, and G. Carleo, Deep autoregressive models for the efficient variational simulation of many-body quantum systems, Phys. p. 0 Rev. Lett. 124, 020503 (2020) <ref:2604. p. 006741#pg4>
[12] M. Hibat-Allah, M Ganahl, R Melko and J Carrasquilla, Recurrent neural network wave functions, Phys. p. 0 Rev. Research 2, 023358 (2020) <ref:2604. p. 006741#pg4>
[13] J Carrasquilla, Machine learning for quantum matter, Advances in Physics: X 5, 1797528 (2020) <ref:2604.06741#pg4>
[14] C. Vivas and J Madronero, Neural-Network Quantum States: A Systematic Review, arXiv preprint (2022), arXiv:2204. p. 012966 <ref:2604. p. 006741#pg4>
[15] H. Lange, A Van de Walle, A Abedinnia and A Bohrdt, From architectures to applications: a review of neural quantum states, Quantum Science and Technology 9, 040501 (2024) <ref:2604. p. 006741#pg4>
[16] P. Chen, B. Yan, and S. X. Cui, Representing arbitrary ground states of the toric code by a restricted Boltzmann machine, Phys. p. 0 Rev. B 111, 045101 (2025) <ref:2604. p. 006741#pg4>
[17] N Paul, Bound on entanglement in neural quantum states (2025), arXiv:2510.11797 [quant-ph] <ref:2604. p. 006741#pg4>
[18] X Chen, Z.-C Gu, and X.-G Wen, Complete classification of one-dimensional gapped quantum phases in interacting spin systems, Phys. p. 0 Rev. B 84, 235128 (2011) <ref:2604. p. 006741#pg4>
[19] N Schuch, D Perez-García and I Cirac, Classifying quantum phases using matrix product states and projected entangled pair states, Phys. Rev. B 84, 165139 (2011) <ref:2604. p. 006741#pg4>
[20] F Pollmann, A M Turner, E Berg and M Oshikawa, Symmetry protection of topological phases in one-dimensional quantum spin systems, Phys. Rev. B 85, 075125 (2012) <ref:2604. p. 006741#pg4>
[21] J H Han, E Lake, H T Lam, R Verresen and Y You, Topological quantum chains protected by dipolar and other modulated symmetries, Phys. Rev. B 109, 125121 (2024) <ref:2604. p. 006741#pg4>
[22] H T Lam, Classification of dipolar symmetry-protected topological phases: Matrix product states, stabilizer Hamiltonians, and finite tensor gauge theories, Phys. Rev. B 109, 115142 (2024) <ref:2604. p. 006741#pg4>
[23] J Kim, Y You and J H Han, Noninvertible symmetry and topological holography for modulated SPT in one dimension, SciPost Phys. 19, 110 (2025) <ref:2604. p. 006741#pg4>
[24] S Sun, J-H Zhang, Z Bi and Y You, Holographic View of Mixed-State Symmetry-Protected Topological Phases in Open Quantum Systems, PRX Quantum 6, 020333 (2025) <ref:2604.06741#pg4>
[25] S D Pace, G Delfino, H T Lam and Omer M Aksoy, ¨ Gauging modulated symmetries: Kramers-Wannier dualities and non-invertible reflections, SciPost Phys. 18, 021 (2025) <ref:2604. p. 006741#pg4>
[26] T Saito, W Cao, B Han and H Ebisu, Matrix product state classification of one-dimensional multipole symmetryprotected topological phases, Physical Review B 112, 10.1103/6txg-7hy7 (2025) <ref:2604. p. 006741#pg4>
[27] D Bulmash, Defect Networks for Topological Phases Protected By Modulated Symmetries (2025), arXiv:2508.06604 <ref:2604. p. 006741#pg4>
[28] S D Pace, Omer M Aksoy and H T Lam, Spacetime symmetry-enriched SymTFT: From LSM anomalies to modulated symmetries and beyond, SciPost Phys. 20, 007 (2026) <ref:2604. p. 006741#pg4>
Appendix A Details
The proof of symmetry fractionalization in the cluster state is shown for the open-chain ground state by applying identities successively to obtain a final form <ref:2604.06741#pg4>. The proof of symmetry fractionalization proceeds similarly for the P-representation, where acting on the open-chain ground state with C1 yields a specific transformation <ref:2604. p. 306741#pg4>.
Appendix B Details
The derivation of the weight matrix W(s,h) for the ZN cluster state starts from an ansatz and involves finding an integer r satisfying certain modular arithmetic conditions to obtain the RBM representation <ref:2604.06741#pg4>. The conditions for the periodicity of the summand in Eq. p. 1 (B4) are specified regarding 2a and 4a depending on whether N is odd or even <ref:2604. p. 106741#pg4>.
Appendix C Details
The proof of symmetry invariance for the MPS representation in Appendix C is much more involved than the proof made in the NQS representation <ref:2604.06741#pg4>. The transformation of the MPS matrix under C†1 involves several identities to show that the overall wavefunction stays invariant <ref:2604. p. 506741#pg4>.
Appendix D Details
The proof of symmetry for dCSII under C1, C2, D1, and D2 is achieved by summing over hidden variables to recover the original wavefunction in Eq. (D5) <ref:2604. p. 006741#pg4>. The transformation of the P-representation under C†2 is shown through a sequence of index relabelings and property applications <ref:2604.06741#pg4>.
The local MPO representation of Kc can be derived by combining two traces, which can be re-written as a single trace involving products of projectors and weight matrices <ref:2604.06741#pg4>. The local tensor (K j c) is defined as a product of terms involving W and W† matrices <ref:2604.06741#pg4>.
The type-II dipolar cluster model admits a noninvertible symmetry Kd which is shown to be equivalent to a quadrupole Fourier transformation <ref:2604.06741#pg4>. The Kd-invariance of the dCSII ground state immediately implies the first four fusion rules satisfied by the symmetry generators <ref:2604.06741#pg4>. The MPO representation of Kd is given in terms of TPS representations, where each bond index is a collection of four ZN numbers <ref:2604.06741#pg4>. The Kd-invariance proof for dCSII shows that the image lies entirely in the sector C1 = C2 = D1 = D2 = 1 <ref:2604.06741#pg4>.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Improved AI can construct neural quantum states (NQS) for ZN cluster and dipolar cluster states using a projector-based scheme called the P-representation, which is an
efficient way to express cluster and dipolar cluster state’s wavefunctions.
-
Improved AI can develop new representations like the matrix product state (MPS) and tensor product state (TPS) from the NQS by re-writing the interaction matrix in terms of neural weight matrices using a restricted Boltzmann machine scheme.
-
Improved AI can derive closed form weight functions W(s,h) that couple physical spins to hidden variables, generalizing previous constructions for Z2 cluster states to ZN.
-
Improved AI can analyze and prove symmetry invariance for cluster states by performing manipulations in the P-representation, demonstrating how
the same analysis can be performed just as easily, if not more so, using the Prepresentation.
-
Improved AI can construct explicit NQS representations for ZN cluster states and two types of dipole-conserving ZN cluster states (type-I and type-II), deriving closed-form interaction weight matrices for ZN variables.
-
Improved AI can determine whether a state naturally requires the TPS representation by benchmarking the resulting TPS construction against conventional MPS representation using density-matrix renormalization group (DMRG) simulations, arguing that
the TPS could offer a more efficient representation for some modulated SPT states.
-
Improved AI can interpret the KramersWannier (KW) operator as
the dipolar generalization of the discrete Fourier transform on ZN variables,
providing intuition for why the KW map is non-invertible becausethe target variable dj ≡ sj − sj+1 has a constraint, ∑j dj = 0.
-
Improved AI can construct an MPO representation of the KW operator using various methods, including one that involves a
pair of projectors at each site,
which is useful for analyzing non-invertible symmetries like the Kd operator.
Sources
- Quantum Entanglement in Neural Network States
- Neural-Network Quantum States: A Systematic Review
- Bound on entanglement in neural quantum states
- Classification of Dipolar Symmetry-Protected Topological Phases: Matrix Product States, Stabilizer Hamiltonians and Finite Tensor Gauge Theories
- Defect Networks for Topological Phases Protected By Modulated Symmetries
- Matrix Product States for Modulated Symmetries: SPT, LSM, and Beyond
- Matrix Product States for Modulated Topological Phases: Crystalline Equivalence Principle and Lieb-Schultz-Mattis Constraints
- Lattice Translation Modulated Symmetries and TFTs
- What's Done Cannot Be Undone: TASI Lectures on Non-Invertible Symmetries
- Large Electron Model: A Universal Ground State Predictor
Related papers
- Topological metal-insulator transitions in one-dimensional non-Hermitian quasicrystals: beyond PT-symmetry
- Microscopic and collective signatures of feature learning in neural networks
- Synchronous and asynchronous Monte Carlo dynamics of mean-field spin glasses with rotationally invariant couplings: what they share and where they differ
- Reentrant Localization Length in a Quasiperiodic Thue-Morse Chain
- Beyond Ground States: Learning Many-Body Hamiltonians for p-Bit Invertible Arithmetic
- Linear equivalence of nonlinear recurrent neural networks