On generating direct powers of dynamical Lie algebras

Jonathan Allcock; Miklos Santha; Pei Yuan; Shengyu Zhang

REVIEW 2 major objections 5 minor 1 cited by

On generating direct powers of dynamical Lie algebras

T0 review · 2 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Tensoring every generator of a cyclic dynamical Lie algebra with a single sign-unambiguous Hermitian operator clones the nonabelian part into K independent copies, adding no generators and only ⌈log K⌉ qubits.

desk verdict Core direct-power constructions are new and mostly correct; the QAOA application is misstated in Theorem 14 and the abstract overclaims for cyclic DLAs with nontrivial center. read the letter →

arxiv 2506.05733 v1 pith:HD2XJUE3 submitted 2025-06-06 quant-ph math-phmath.MP

Jonathan Allcock , Miklos Santha , Pei Yuan , Shengyu Zhang This is my paper

classification quant-phmath-phmath.MP MSC 17B0581P68

keywords dynamicalLiealgebradirectsumparameterizedquantumcircuitsbarrenplateauscyclicgeneratingsetQAOA-MaxCutPaulitensorproductconstruction

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks what happens to the dynamical Lie algebra (DLA) of a parameterized quantum circuit when every generator is tensored with a single fixed Hermitian operator. Its central claim is that this modification can make the new DLA isomorphic to a direct sum of K copies of the original one, using only about log K additional qubits and, for a broad class of 'cyclic' generator sets, no additional parameters at all. Because the dimension of the DLA controls both how expressive a circuit is and how vulnerable it is to barren plateaus, this gives a concrete way to design circuits whose reachable algebra is a controllable multiple of a small seed algebra. The paper proves a general tensor construction and then isolates the cyclic case, where the copy count K is achieved exactly at no extra generator cost.

What carries the argument

The proof mechanism is a Vandermonde argument over the eigenspaces of $Q$. Because $Q$ has $K$ distinct eigenvalues, its powers $Q^0, Q^1, \dots, Q^{K-1}$ are linearly independent and span the same space as the spectral projectors $\Pi_1, \dots, \Pi_K$. The cyclic property guarantees that every basis element $V$ of $[\mathfrak{g}_A, \mathfrak{g}_A]$ can be written as a nested commutator in which every generator carries a factor of $Q$, so the value is $V \otimes Q^\ell$; inserting stable extensions adds $M$ further factors, giving $V \otimes Q^{\ell + jM}$ for all $j$. Sign-unambiguousity makes the powers $Q^{\ell+jM}$ distinct scalars on each eigenvalue sector, so the relevant Vandermonde matrix is nonsingular and the span of these powers is the span of the projectors. Lemma 9 then maps $V \otimes \Pi_j$ to the $j$-th copy of $[\mathfrak{g}_A, \mathfrak{g}_A]$.

What would settle it

Choose a sign-unambiguous $Q$ with three distinct eigenvalues (for example, the two-qubit diagonal operator $\mathrm{diag}(1,2,3)$), take the QAOA-MaxCut generators $M_G = \{A_1, A_2\}$ for a small graph such as a triangle, and compute the dimension of $[\mathfrak{g}_{M_{G,Q}}, \mathfrak{g}_{M_{G,Q}}]$ explicitly. If Theorem 5's premise holds, this dimension should equal $3 \dim [\mathfrak{g}_{M_G}, \mathfrak{g}_{M_G}]$; finding a smaller dimension would show the tensored set is not cyclic for three-eigenvalue $Q$, restricting the no-new-parameter claim to two-eigenvalue observables.

Watch

Extended reading notes

Core claim

The central discovery is that a tensor factor can multiply the nonabelian part of a dynamical Lie algebra. Given a dynamical generating set $A$ and a Hermitian operator $\chi$ with $K$ distinct eigenvalues, the set $A' = \{A_i \otimes \chi^j : i \in [L], 0 \le j \le K-1\}$ generates $\bigoplus_{j=1}^K \mathfrak{g}_A$ (Theorem 1), and the smaller set $\{A_i \otimes I\} \cup \{A_i \otimes \chi\}$ already generates $\bigoplus_{j=1}^K [\mathfrak{g}_A, \mathfrak{g}_A]$ with a center of dimension $2\dim Z(\mathfrak{g}_A)$ (Theorem 2). The sharper cardinality-preserving result (Theorem 5) requires the generating set to be cyclic and $Q$ to be sign unambiguous with $K$ distinct nonzero eigenvalues. Then $A_Q = \{A_i \otimes Q\}$ satisfies $[\mathfrak{g}_{A_Q}, \mathfrak{g}_{A_Q}] \cong \bigoplus_{j=1}^K [\mathfrak{g}_A, \mathfrak{g}_A]$ and $Z(\mathfrak{g}_{A_Q}) = \{C \otimes Q : C \in Z(\mathfrak{g}_A)\}$. In words, the commutator part of the algebra splits into $K$ independent copies, one per eigenspace of $Q$, while the center rides along unchanged through $Q$.

Load-bearing premise

The whole no-new-parameter construction rests on the 'cyclic' property of the generating set: every noncommuting pair of generators must admit a nested commutator, built from the generators, that returns the pair's commutator up to a scalar; for the QAOA-MaxCut application this property is asserted after explicit computation of a few commutators, so that computation is the premise most worth checking independently.

Editorial extensions

If this is right

For any cyclic generating set—which the paper shows includes Pauli DLAs, QAOA-MaxCut DLAs, and two-generator sets where one generator squares to a scalar—any sign-unambiguous $Q$ with $K$ eigenvalues yields a new circuit whose nonabelian reachable algebra is $K$ independent copies of the original, using the original number of parameters and $\lceil \log K \rceil$ extra qubits.
The direct-power structure splits the loss-function variance bound into a sum over $K$ copies, so the variance picks up a factor of $1/K$ relative to the single-copy case; this gives a parameter-preserving knob for suppressing barren-plateau decay.
For non-cyclic sets, the same goal is still achievable by the cardinality-increasing constructions of Theorems 1 and 2, at the cost of a factor $K$ or $2$ in the number of generators; the paper shows this extra cost is unavoidable in the worst case because the center of a DLA has dimension at most $|A|$.
The constructions produce explicit bases for the $K$ copies via the spectral projectors of the tensored operator, so the isomorphism is not merely abstract: one can identify which linear combinations of nested commutators live in which copy.

Reading between the lines

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

The sign-unambiguous condition is likely stronger than necessary: the Vandermonde step only needs the eigenvalues of $Q^M$ to be distinct, so a $Q$ whose eigenvalue ratios are not $M$-th roots of unity would also work; relaxing this condition is a direct and easy test of the method's limits.
The same tensor-with-an-observable trick can be iterated, tensoring first with $Q_1$ and then with $Q_2$ on the new register, to produce hierarchical direct products of DLAs; such nested structures might model layered variational ansätze whose reachable algebras grow multiplicatively with depth.
The result reframes the expressibility/trainability trade-off: instead of choosing between a small polynomial-dimension DLA and a large exponential one, one can start from a small seed algebra and multiply its dimension by $K$ at logarithmic qubit cost, making the trade-off continuously tunable.
Because Theorem 15's square-to-scalar condition covers many hardware-native gate sets (involutions) beyond Pauli strings, the no-new-parameter construction may apply broadly to common ansatz families; checking which practical generator sets are cyclic is a natural next step.

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

The useful core here is real. Theorems 1, 2, and 5 give a genuine toolkit for engineering DLAs that are direct sums of copies of [g_A,g_A] using only log K extra qubits, and Theorem 5's cardinality-preserving version is the kind of result people working on barren-plateau avoidance and overparameterization will want to cite. The proofs of these theorems are detailed and I followed the induction and Vandermonde arguments without finding gaps. Theorem 3's connection to Aguilar et al.'s Pauli classification is also a nice conceptual link, not just a repackaging.

The soft spots are real but mostly editorial. Theorem 14 as stated is false: M_{G,Q} and S_{G,Q} are not cyclic when Q has more than one distinct eigenvalue. Any nested commutator of tensored generators carries Q^m, and a stable extension would need Q^m proportional to Q^2, which fails for sign-unambiguous Q with K>1 and m≥3. The displayed identities are clearly meant for the untensored sets M_G and S_G, so the fix is to restate the theorem and then apply Theorem 5 with A=M_G (or S_G). I checked the stress-test's attempted counterexample for S_G: it does not hold up, because A3^2=-I, so Theorem 15 gives [A3,[A3,[A3,A1]]] proportional to [A3,A1], not zero. But the stress-test is right that the proof is only 'by explicitly computing'—for a claim that carries the whole QAOA-MaxCut application, that is too thin. A referee should ask for a short lemma with the actual computation for arbitrary graphs.

The abstract also overclaims. For cyclic DLAs with nontrivial center, Theorem 5 produces K copies of [g_A,g_A], not K copies of g_A; the center becomes C⊗Q, a single copy. The main text is careful about this, but the abstract is not. Minor: the remark after Theorem 2 has a typo in the center dimension formula.

Overall, the central theorems are solid and worth engaging with. The paper deserves a serious referee; the fixes are small but necessary: restate Theorem 14, supply the missing cyclicity computation, and qualify the abstract. I would take it to a reading group and would cite Theorem 5 in my own work.

Referee Report

2 major / 5 minor

Summary. The paper studies how to modify a dynamical generating set A so that the generated DLA becomes a direct sum of K copies of the original DLA (or of its commutator subalgebra), using few additional qubits and few additional generators. Theorem 1 uses the K powers of a Hermitian operator χ; Theorem 2 uses only {I,χ} and doubles the generator count while producing K copies of the commutator part; Theorems 3 and 4 specialize to Pauli and two-generator DLAs; Theorem 5 gives a cardinality-preserving construction for cyclic generating sets using a sign-unambiguous Q. Applications to Pauli DLAs (Theorem 16) and QAOA-MaxCut DLAs (Theorem 14) are claimed. The proofs of Theorems 1-5 are mostly detailed inductions with Vandermonde arguments and appear correct. The main weakness is the statement and proof of Theorem 14, which is the only support for the QAOA-MaxCut application.

Significance. If the results are correct, they provide a clean toolkit for engineering DLAs that are direct powers of a given algebra, with logarithmic qubit overhead and, in the cyclic case, no increase in generator count. This is directly relevant to barren-plateau avoidance and overparameterization in variational quantum algorithms. The central Theorem 5 is elegant, and the paper's proofs of Theorems 1-5 are careful and largely self-contained, relying on explicit inductions, Jacobi identities, and Vandermonde determinants rather than on external machinery. The paper does not provide machine-checked proofs, but the arguments are checkable. The QAOA-MaxCut application is currently not established as stated, which is a significant gap in a headline claim; the gap appears fixable, but it must be addressed before the paper can be accepted.

major comments (2)

[Section 5, Theorem 14] Theorem 14 as stated is false for K>1. For a generating set A_Q = {A⊗Q}, every nested commutator built from r factors carries a factor Q^r. A stable extension of the ordered pair (A_i⊗Q, A_j⊗Q) would therefore have to satisfy Q^{r+2} ∝ Q^2 for some extension length r≥1, i.e. Q^r ∝ I. This is impossible for a sign-unambiguous Q with K>1 distinct nonzero eigenvalues, because λ_i^r = λ_j^r would force λ_i = ±λ_j, and the sign-unambiguous condition forbids λ_i = -λ_j. The displayed identities in the proof, such as [A1,[A1,[A1,A2]]] ∝ [A1,A2], are identities for the untensored generators A1,A2; with A1 replaced by A1⊗Q, the left-hand side carries Q^4 while the right-hand side carries Q^2. The intended statement must be that the untensored sets M_G and S_G are cyclic, not M_{G,Q} and S_{G,Q}. This correction is load-bearing because Theorem 5 requires cyclicity of A, and the abstract's claim that QAOA-MaxCut DLAs fall under the cardinality-preserving construction relies on it.
[Section 5, Theorem 14 (proof)] The proof asserts cyclicity of the (corrected) untensored sets by the phrase "by explicitly computing the appropriate nested commutators", but no computation or derivation is given for arbitrary graphs; the displayed identities are merely listed. Since Theorem 5's cyclicity hypothesis is the only point at which this property enters, the QAOA-MaxCut application is currently unsupported. The authors should supply a lemma with a complete proof that for every graph G the identities [A1,[A1,[A1,A2]]] ∝ [A1,A2], [A1,[A1,[A1,A3]]] ∝ [A1,A3], [A2,[A2,[A2,A3]]] ∝ [A2,A3], and [A2,[A2,[A1,A2]]] ∝ [A1,A2] hold. They should also state explicitly that an identity [Ai,[Ai,[Ai,Aj]]] ∝ [Ai,Aj] is a stable extension for both orderings of the pair, because [Ai,Aj] and [Aj,Ai] are proportional up to sign; this addresses any orientation concern. A direct check for a single-edge two-qubit graph shows that the relevant nested commutators are nonzero multiples, so the issue is not a counterexample but a missing general derivation.

minor comments (5)

[Section 4, remark after Theorem 2] The displayed formula "dim(Z(g_{A'_q})) = q dim(Z(g_{A'_q}))" is a typo; it should read "dim(Z(g_{A'_q})) = q dim(Z(g_A))".
[Section 5, Theorem 5 proof] In the center-inclusion part of the proof, "for every C ∈ Z(g_{A_Q})" should be "for every C ∈ Z(g_A)", since the claim is that C⊗Q belongs to the center for every central C of the original algebra.
[Section 5, Theorem 14] The proof writes "in g_{M_{G,Q}} we have ..." and then displays identities using A1, A2, A3 without indicating that these are the untensored generators; the authors should first prove the identities for M_G and S_G and then invoke Theorem 5 for A_Q.
[Section 5, definition of common cycle length] The definition of e(i,j) as the length of a stable extension is ambiguous: it should specify that e(i,j) is the length of the appended sequence, not the length of the full concatenated sequence, and a short example would help the reader.
[Abstract and Section 2] The text uses "log K" in the abstract and introduction but "⌈log K⌉" in the theorems; the notation should be made uniform.

Circularity Check

0 steps flagged · score 1.0 of 10

No load-bearing circularity: the direct-power theorems are proved from definitions and standard Lie-algebra facts, with only a non-load-bearing self-citation.

full rationale

Walking the claimed derivation chain, I find no circular step. Theorems 1, 2, 3, 4, and 5 are established from the definitions of dynamical generating sets, right-nested commutator bases, spectral decompositions of chi and Q, the Vandermonde determinant, and standard Jacobi-identity arguments. The cyclic hypothesis of Theorem 5 is a concrete, checkable property of the generator set, not a restatement of the conclusion; the proof uses a stable extension to obtain V tensor Q^(ell+jM) and then reconstructs the spectral projectors of Q via the sign-unambiguous condition, without assuming the desired isomorphism. The only direct citation of the authors' prior work is the bound dim(Z(g_A)) <= |A| from reference [9], used solely to argue that the factor-K generator count in Theorem 1 is worst-case necessary; the positive construction does not rely on this bound, so the citation is not load-bearing. There are no fitted constants, no prediction that reduces to its input, and no uniqueness claim imported from the authors' other papers. Separately, as a correctness risk rather than as circularity, Theorem 14's proof consists of the phrase 'by explicitly computing the appropriate nested commutators,' and the displayed identities appear to be written for the untensored generators: in the tensored algebra the analogous commutator carries Q^4 on one side and Q^2 on the other, which is not proportional for a sign-unambiguous Q with more than two distinct eigenvalues. This is a gap in the QAOA-MaxCut application, not an instance of self-referential derivation.

Assumptions & free parameters 0 free parameters · 6 assumptions · 0 invented entities

The central constructions depend on standard Lie algebra facts (reductive decomposition, right-nested commutator bases), the defining assumptions of dynamical generating sets, and the new cyclic condition and sign-unambiguous condition introduced to make the cardinality-preserving construction work. No free parameters or invented entities are introduced.

assumptions (6)

standard math Every DLA as a subalgebra of su(2^n) is reductive, i.e., g = [g,g] + Z(g) (Fact 2).
Cited to Knapp [13]; used in Lemma 13 and Theorem 4 center computations.
standard math All nested commutators are linear combinations of right-nested commutators (Fact 1).
Proved in Appendix A via Jacobi identity; used to choose right-nested bases in Lemma 13 and Theorem 4.
domain assumption The generators in A form a dynamical generating set: linearly independent, traceless, anti-Hermitian operators on n qubits.
Baseline assumption throughout; ensures the DLA is a reductive subalgebra of su(2^n).
domain assumption For Pauli DLAs, every generator squares to -I (eigenvalues +/- i), so [A,[A,[A,B]]] = -4[A,B], making the set cyclic with common cycle length 2.
Used in Theorem 16 to show Pauli sets are cyclic.
domain assumption The QAOA-MaxCut generators satisfy stable-extension identities such as [A1,[A1,[A1,A2]]] proportional to [A1,A2] (Theorem 14), verified by explicit computation.
Load-bearing for applying Theorem 5 to QAOA-MaxCut; the full computation is not shown in the text.
domain assumption Q is sign unambiguous and has K distinct nonzero eigenvalues.
Technical condition needed to ensure the powers Q^(l+jM) form an invertible Vandermonde system; without it the cyclic construction may fail.

how reviews work

0 comments

Cite this review

Pith. "Pith review of On generating direct powers of dynamical Lie algebras." pith.science (2026). https://pith.science/paper/HD2XJUE3

@misc{pith2026250605733,
  author       = {Pith},
  title        = {Pith review of: On generating direct powers of dynamical Lie algebras},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HD2XJUE3}},
  note         = {Machine review of arXiv:2506.05733}
}

abstract

The expressibility and trainability of parameterized quantum circuits has been shown to be intimately related to their associated dynamical Lie algebras (DLAs). From a quantum algorithm design perspective, given a set $A$ of DLA generators, two natural questions arise: (i) what is the DLA $\mathfrak{g}_{A}$ generated by ${A}$; and (ii) how does modifying the generator set lead to changes in the resulting DLA. While the first question has been the subject of significant attention, much less has been done regarding the second. In this work we focus on the second question, and show how modifying ${A}$ can result in a generator set ${A}'$ such that $\mathfrak{g}_{{A}'}\cong \bigoplus_{j=1}^{K}\mathfrak{g}_{A}$, for some $K \ge 1$. In other words, one generates the direct sum of $K$ copies of the original DLA. In particular, we give qubit- and parameter-efficient ways of achieving this, using only $\log K$ additional qubits, and only a constant factor increase in the number of DLA generators. For cyclic DLAs, which include Pauli DLAs and QAOA-MaxCut DLAs as special cases, this can be done with $\log K $ additional qubits and the same number of DLA generators as ${A}$.

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

A Dynamical Lie-Algebraic Framework for Hamiltonian Engineering and Quantum Control
quant-ph 2026-03 reject novelty 5.0 of 10

A Lie-algebra toolkit that composes, preserves, and reduces Hamiltonian generator sets, including a nearest-neighbor su(2^N) generating set and a filtering-operator reduction, though the reduction proof and one error-...

Reference graph

Works this paper leans on

15 extracted references · 10 canonical work pages · cited by 1 Pith paper

[1]

A Lie algebraic theory of barren plateaus for deep param- eterized quantum circuits.Nature Communications, 15(1):7172, 2024

Michael Ragone, Bojko N Bakalov, Frédéric Sauvage, Alexander F Kemper, Carlos Ortiz Mar- rero, Martín Larocca, and M Cerezo. A Lie algebraic theory of barren plateaus for deep param- eterized quantum circuits.Nature Communications, 15(1):7172, 2024

work page 2024
[2]

Characterizing barren plateaus in quantum ansätze with the adjoint representation.Nature Communications, 15(1):7171, 2024

Enrico Fontana, Dylan Herman, Shouvanik Chakrabarti, Niraj Kumar, Romina Yalovetzky, Jamie Heredge, Shree Hari Sureshbabu, and Marco Pistoia. Characterizing barren plateaus in quantum ansätze with the adjoint representation.Nature Communications, 15(1):7171, 2024

work page 2024
[3]

Theory of overparametrization in quantum neural networks.Nature Computational Science, 3(6):542–551, 2023

Martin Larocca, Nathan Ju, Diego García-Martín, Patrick J Coles, and Marco Cerezo. Theory of overparametrization in quantum neural networks.Nature Computational Science, 3(6):542–551, 2023

work page 2023
[4]

Cambridge university press, 1997

Velimir Jurdjevic.Geometric control theory. Cambridge university press, 1997

work page 1997
[5]

Full classification of Pauli Lie algebras

Gerard Aguilar, Simon Cichy, Jens Eisert, and Lennart Bittel. Full classification of Pauli Lie algebras. arXiv preprint arXiv:2408.00081, 2024

arXiv 2024
[6]

A quantum approximate optimization algorithm

Edward Farhi, Jeffrey Goldstone, and Sam Gutmann. A quantum approximate optimization algorithm. arXiv preprint arXiv:1411.4028, 2014. 12

arXiv 2014
[7]

Classification of dynamical Lie algebras generated by spin interactions on undirected graphs.arXiv preprint arXiv:2409.19797, 2024

Efekan Kökcü, Roeland Wiersema, Alexander F Kemper, and Bojko N Bakalov. Classification of dynamical Lie algebras generated by spin interactions on undirected graphs.arXiv preprint arXiv:2409.19797, 2024

arXiv 2024
[8]

Analyzing the quantum approximate optimization algorithm: ansätze, symmetries, and Lie algebras

Sujay Kazi, Martín Larocca, Marco Farinati, Patrick J Coles, M Cerezo, and Robert Zeier. Analyzing the quantum approximate optimization algorithm: ansätze, symmetries, and Lie algebras. arXiv preprint arXiv:2410.05187, 2024

arXiv 2024

Show all 15 references

[9]

On the dynamical Lie algebras of quantum approximate optimization algorithms.arXiv preprint arXiv:2407.12587, 2024

Jonathan Allcock, Miklos Santha, Pei Yuan, and Shengyu Zhang. On the dynamical Lie algebras of quantum approximate optimization algorithms.arXiv preprint arXiv:2407.12587, 2024

2024 arXiv
[10]

Controllability of the periodic quantum Ising spin chain and the Onsager algebra

Domenico D’Alessandro and Yasemin Isik. Controllability of the periodic quantum Ising spin chain and the Onsager algebra. Journal of Physics A: Mathematical and Theoretical, 58(11):115202, 2025

2025
[11]

Symmetry criteria for quantum simulability of effective interactions

Zoltán Zimborás, Robert Zeier, Thomas Schulte-Herbrüggen, and Daniel Burgarth. Symmetry criteria for quantum simulability of effective interactions. Physical Review A, 92(4):042309, 2015

2015
[12]

A lie algebraic theory of barren plateaus for deep param- eterized quantum circuits.Nature Communications, 15(1):7172, 2024

Michael Ragone, Bojko N Bakalov, Frédéric Sauvage, Alexander F Kemper, Carlos Ortiz Mar- rero, Martín Larocca, and M Cerezo. A lie algebraic theory of barren plateaus for deep param- eterized quantum circuits.Nature Communications, 15(1):7172, 2024

2024
[13]

Springer, 1996

Anthony W Knapp.Lie groups beyond an introduction, volume 140. Springer, 1996

1996
[14]

Controllability of symmetric spin networks

Francesca Albertini and Domenico D’Alessandro. Controllability of symmetric spin networks. Journal of Mathematical Physics, 59(5), 2018

2018
[15]

The- oretical guarantees for permutation-equivariant quantum neural networks.npj Quantum Infor- mation, 10(1):12, 2024

Louis Schatzki, Martin Larocca, Quynh T Nguyen, Frederic Sauvage, and Marco Cerezo. The- oretical guarantees for permutation-equivariant quantum neural networks.npj Quantum Infor- mation, 10(1):12, 2024. A Proof of Fact 1 Here we give a proof of Fact 1 from the main text. Let ...

2024

Pith tools

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

[1] [1]

A Lie algebraic theory of barren plateaus for deep param- eterized quantum circuits.Nature Communications, 15(1):7172, 2024

Michael Ragone, Bojko N Bakalov, Frédéric Sauvage, Alexander F Kemper, Carlos Ortiz Mar- rero, Martín Larocca, and M Cerezo. A Lie algebraic theory of barren plateaus for deep param- eterized quantum circuits.Nature Communications, 15(1):7172, 2024

work page 2024

[2] [2]

Characterizing barren plateaus in quantum ansätze with the adjoint representation.Nature Communications, 15(1):7171, 2024

Enrico Fontana, Dylan Herman, Shouvanik Chakrabarti, Niraj Kumar, Romina Yalovetzky, Jamie Heredge, Shree Hari Sureshbabu, and Marco Pistoia. Characterizing barren plateaus in quantum ansätze with the adjoint representation.Nature Communications, 15(1):7171, 2024

work page 2024

[3] [3]

Theory of overparametrization in quantum neural networks.Nature Computational Science, 3(6):542–551, 2023

Martin Larocca, Nathan Ju, Diego García-Martín, Patrick J Coles, and Marco Cerezo. Theory of overparametrization in quantum neural networks.Nature Computational Science, 3(6):542–551, 2023

work page 2023

[4] [4]

Cambridge university press, 1997

Velimir Jurdjevic.Geometric control theory. Cambridge university press, 1997

work page 1997

[5] [5]

Full classification of Pauli Lie algebras

Gerard Aguilar, Simon Cichy, Jens Eisert, and Lennart Bittel. Full classification of Pauli Lie algebras. arXiv preprint arXiv:2408.00081, 2024

arXiv 2024

[6] [6]

A quantum approximate optimization algorithm

Edward Farhi, Jeffrey Goldstone, and Sam Gutmann. A quantum approximate optimization algorithm. arXiv preprint arXiv:1411.4028, 2014. 12

arXiv 2014

[7] [7]

Classification of dynamical Lie algebras generated by spin interactions on undirected graphs.arXiv preprint arXiv:2409.19797, 2024

Efekan Kökcü, Roeland Wiersema, Alexander F Kemper, and Bojko N Bakalov. Classification of dynamical Lie algebras generated by spin interactions on undirected graphs.arXiv preprint arXiv:2409.19797, 2024

arXiv 2024

[8] [8]

Analyzing the quantum approximate optimization algorithm: ansätze, symmetries, and Lie algebras

Sujay Kazi, Martín Larocca, Marco Farinati, Patrick J Coles, M Cerezo, and Robert Zeier. Analyzing the quantum approximate optimization algorithm: ansätze, symmetries, and Lie algebras. arXiv preprint arXiv:2410.05187, 2024

arXiv 2024