Pith.

Quantum Physics · arXiv:2611.16791 v1 · submitted 12 Nov 2026 · 38 references

Non-Convex Optimization via Quantum Tunneling Search on a Photonic Chip

Xi-Jun Yuan, Shuo Zhou, Yizhou Liu, Xuan Chen, Hong-Jing Du, Xiao-Wen Shang, Wen-Hao Zhou, Yu-Xuan Fu

show 10 more authors

Bao-Jing Liu, Ze-Feng Lan, Hang Song, Fang-Song Nie, Yu-Lin Yu, Xiao-Yun Xu, Jian-Peng Dou, Tongyang Li, Hao Tang, Xian-Min Jin

The paper says

“Non-convex optimization plays an important role in modern machine learning since plenty of real-world problems have objective functions or constraints with non-convex properties.”

more of the abstract

The non-convex nature of landscapes gives rise to numerous local minima and saddle points, dramatically hindering the ability of gradient descent-based methods to explore the landscape and obtain global optima. Here we present quantum tunneling search (QTS), an algorithm based on quantum tunneling walk and local annealing, which effectively explores the non-convex landscape by leveraging the quantum tunneling effect. In experiments, we map an objective function to the potential term of the waveguide Hamiltonian by controlling the refractive index with the femtosecond direct laser writing technique. With a proper input state, the average probability distributions on 10-micrometer-spaced 29 x 1 and 21 x 9 waveguide arrays highlight the local minima of various 1D and 2D non-convex landscapes, respectively. By further introducing the varying Hamiltonian approach, more efficient search and even the identification of the global minimum are achieved. Our work experimentally verifies that QTS is a successful and efficient non-convex optimization solver, demonstrating quantum speedups over classical stochastic gradient descent (SGD), which serves as a practical application of near-term analog quantum simulators in the field of machine learning.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

The passes disagreed. Reader: holds up, with conditions. Skeptic: does not hold up. Referee: major revision.

5 major objections · 7 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 38 references say.

The paper says, at Abstract, §II (Multi-well random landscape), Supplementary Note 1 (Benchmark accounting and scope)

The review did not quote this passage.

A classifier read this sentence of the paper as acknowledging the concern (probability 0.91):

“The data presented here do not allow such an end-to-end speedup estimate.”

Pith says: Objection 1 of 5 · major

The claim of quantum speedups over SGD (sentences b5, b23) is not supported by the data. The supplementary material explicitly states that the horizontal axes are fundamentally different (propagation length vs. iterations), the plotting factor of 60 is a display convention and not an independently established comparison of resources, and the data presented do not allow an end-to-end speedup estimate (sentences b180, b181, b184).

more of objection 1

The SGD baseline is a simple gradient descent with noise, not a state-of-the-art classical optimizer. The theoretical speedup proven in Ref. [5] assumes access to the local ground state of the initial well, which the paper attempts to approximate via local annealing, but the quality of this approximation is not rigorously characterized. The paper also acknowledges that the varying Hamiltonian approach does not by itself cause relaxation to the instantaneous ground state (sentence b268). Therefore, the experimental results do not demonstrate quantum speedups, and the claim is not supported by the evidence provided.

The paper says, at §II (Multi-well random landscape), Supplementary Note 1 (Benchmark accounting and scope)

The review did not quote this passage.

A classifier read this sentence of the paper as acknowledging the concern (probability 0.94):

“The plotting factor of 60 is a display convention and not an independently established comparison of resources.”

Pith says: Objection 2 of 5 · major

The comparison to SGD is not a fair resource comparison. The SGD baseline uses a simple gradient descent with noise, not a state-of-the-art classical optimizer. The number of SGD steps is rescaled by a factor of 60 to bring its spatial spread close to that of the photonic search, but this factor is arbitrary and not based on any resource equivalence.

more of objection 2

The paper acknowledges that the data do not allow an end-to-end speedup estimate (sentence b184). Without a meaningful comparison, the claim of quantum speedups is unsubstantiated.

The paper says, at §II (Global minimum search), Supplementary Note 2 (Evidence and controls for tunneling-inspired transport)

The review did not quote this passage.

A classifier read this sentence of the paper as acknowledging the concern (probability 0.95):

“Consequently, output intensity at a distant minimum is consistent with the engineered Hamiltonian dynamics but, on its own, does not isolate the contribution of a programmed potential barrier.”

Pith says: Objection 3 of 5 · major

The experimental results are consistent with the engineered Hamiltonian dynamics, but the paper does not provide controls to isolate the contribution of quantum tunneling. The authors acknowledge that output intensity at a distant minimum is consistent with the engineered Hamiltonian dynamics but does not isolate the contribution of a programmed potential barrier (sentence b232).

more of objection 3

No additional barrier-control data are reported (sentence b234). Without such controls, the claim that the observed behavior is due to quantum tunneling rather than coherent coupling or other classical effects is not fully supported.

The paper says, at §II (Local annealing), Supplementary Note 1 (Initial state significance)

The review did not quote this passage.

Pith says: Objection 4 of 5 · major

The theoretical speedup in Ref. [5] assumes access to the local ground state of the initial well. The paper uses an incomplete local annealing process to approximate this state, but the quality of this approximation is not rigorously characterized. The energy of the input state is reported as E=0.162, compared to the ground state energy E0=0.034 (Figure 3a), but the impact of this deviation on the theoretical speedup is not analyzed.

more of objection 4

The paper acknowledges that the input state is not the exact ground state (sentence b64), but does not quantify how this affects the query complexity or mixing time.

The paper says, at §II (Varying Hamiltonian), Supplementary Note 2 (Varying the ratio)

The review did not quote this passage.

Pith says: Objection 5 of 5 · major

The varying Hamiltonian approach is introduced to accelerate search and identify the global minimum, but the paper acknowledges that increasing the ratio in a lossless, coherently evolving array does not by itself cause relaxation to the instantaneous ground state (sentence b268).

more of objection 5

The mechanism by which the varying Hamiltonian leads to global minimum identification is not clearly explained or theoretically justified. The experimental results show that the probability at the global minimum increases, but without a clear theoretical understanding, it is difficult to assess whether this is due to the varying Hamiltonian or other factors.

7 minor objections

The paper says, at General

Pith says: minor objection 1 of 7

The paper uses the term 'quantum speedups' but the experimental demonstration does not provide a meaningful comparison of resources. The authors should either remove this claim or provide a more careful discussion of what would be required to demonstrate a speedup.

The paper says, at §II (Multi-well random landscape), Figure 4

Pith says: minor objection 2 of 7

The caption of Figure 4 states that the SGD curve shows the empirical distribution of positions after 1,600 iterations, but the text says the number of SGD steps is rescaled by a factor of 60. This is confusing and should be clarified.

show 5 more

The paper says, at Supplementary Note 1 (Benchmark accounting and scope)

Pith says: minor objection 3 of 7

The supplementary material provides a detailed and honest discussion of the limitations of the comparison to SGD. This is commendable, but it directly contradicts the claims made in the abstract and main text. The authors should reconcile these statements.

The paper says, at §II (Double-well landscape)

Pith says: minor objection 4 of 7

The paper states that z=55 mm denotes the hitting 'time', but the assignment of a first detectable arrival length would require a specified intensity threshold and is limited by the spacing between sampled lengths (sentence b199). This should be acknowledged in the main text.

The paper says, at Supplementary Note 2 (Evidence and controls for tunneling-inspired transport)

Pith says: minor objection 5 of 7

The paper suggests a discriminating control experiment but does not perform it. This is a reasonable suggestion for future work, but the lack of such controls should be acknowledged as a limitation in the main text.

The paper says, at §II (Multi-well landscapes)

Pith says: minor objection 6 of 7

The paper states that the maximum available waveguide length is insufficient for convergence (sentence b73). This is an important limitation that should be discussed more prominently.

The paper says, at Supplementary Note 2 (Fabrication overhead and scope of scalability)

Pith says: minor objection 7 of 7

The paper provides a realistic assessment of the fabrication overhead and scalability challenges. This is good, but it further undermines the claim of practical quantum speedups.

What would settle it

Re-analyze the data without the rescaling factor of 60. Compare the number of objective function evaluations (or gradient calls) required by QTS (including fabrication, calibration, and measurement overhead) to achieve a given success probability on a specific landscape, against a state-of-the-art classical optimizer (e.g., CMA-ES, basin-hopping, or a well-tuned SGD with momentum) under a matched budget. If the classical optimizer matches or exceeds QTS performance, the speedup claim is unsupported.

from the skeptic's reading

How the review went

3 readings by deepseek-v4-flash · 13 Nov 2026

  1. Reader holds up, with conditions confidence high
  2. Skeptic agrees: says it does not hold up 5 s · 18,779 tokens read, 566 written The claim of quantum speedups over SGD is not supported by the data presented, as the authors themselves acknowledge in the supplementary material.
  3. Referee major revision: 5 major objections, 7 minor 12 s · 19,041 tokens read, 1,694 written

Not run on this paper: the editor, the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The strongest objection

The central claim of quantum speedups over SGD (sentences b5, b23) is undermined by the paper's own statements. The supplementary material explicitly states that the horizontal axes are fundamentally different (propagation length vs. iterations), the plotting factor of 60 is a display convention and not an independently established comparison of resources, and the data presented do not allow an end-to-end speedup estimate (sentences b180, b181, b184). The SGD baseline is a simple gradient descent with noise, not a state-of-the-art classical optimizer. The theoretical speedup proven in Ref. [5] assumes access to the local ground state of the initial well, which the paper attempts to approximate via local annealing, but the quality of this approximation is not rigorously characterized. The paper also acknowledges that the varying Hamiltonian approach does not by itself cause relaxation to the instantaneous ground state (sentence b268). Therefore, the experimental results do not demonstrate quantum speedups, and the claim is not supported by the evidence provided.

The reader's verdict is CONDITIONAL, but the stress-test reveals a more fundamental issue: the paper's own supplementary material explicitly states that the data do not allow an end-to-end speedup estimate, and the comparison to SGD is not a fair resource comparison. The authors acknowledge that the horizontal axes are fundamentally different, the plotting factor of 60 is a display convention, and the SGD baseline is weak. The theoretical speedup assumes access to the local ground state, which is not rigorously achieved. The paper also acknowledges that the varying Hamiltonian approach does not cause relaxation to the ground state. Therefore, the central claim of quantum speedups is not supported by the evidence, and the paper should be rejected unless the authors provide a fair comparison and rigorous characterization of the input state. The reader correctly identified the weakest assumption, but the verdict should be stronger given the paper's own admissions.

The referee's report

If the claims were fully supported, this would be a significant experimental demonstration of a quantum algorithm for non-convex optimization, a classically hard problem. The photonic platform is interesting and the experimental techniques are non-trivial. However, the central claim of quantum speedups is not supported by the data presented, as the authors themselves acknowledge in the supplementary material. The comparison to SGD is not a fair resource comparison, the SGD baseline is not state-of-the-art, and the theoretical speedup assumptions are not fully met. The experimental results are consistent with the engineered Hamiltonian dynamics but do not isolate the contribution of quantum tunneling or demonstrate a practical advantage. The paper is well-structured and the experimental work is careful, but the claims are overstated relative to the evidence.

Pith's reading of the paper (a machine's summary, not the authors')

The paper reports an experimental demonstration of a quantum optimization algorithm called quantum tunneling search (QTS) on a photonic chip. The authors encode a non-convex objective function into the refractive index profile of a waveguide array, then launch an optical state prepared by local annealing into the array. As the light propagates, quantum tunneling allows probability amplitude to pass through energy barriers that would trap classical gradient-descent methods. By averaging the output intensity over multiple propagation lengths, the system highlights local minima, and by gradually increasing the potential depth (the varying Hamiltonian technique) it can concentrate probability on the global minimum. The authors claim that this approach demonstrates quantum speedups over classical stochastic gradient descent, though they acknowledge that the comparison is limited because the horizontal axes (propagation length vs. iterations) are fundamentally different and the data do not permit an end-to-end speedup estimate. The work is presented as a proof-of-principle for near-term analog quantum simulators in non-convex optimization.

What carries the argument. The central object is the quantum tunneling walk (QTW) implemented in a photonic waveguide array. The Hamiltonian is $H = \sum_i \beta_i a_i^\dagger a_i + \sum_{i \neq j} C_{i,j} a_i^\dagger a_j$, where the on-site propagation constants $\beta_i$ are set proportional to the objective function values $f(x_i)$ at grid points $x_i$, and $C_{i,j}$ gives the nearest-neighbor coupling. Light injected into the array evolves according to $|\Psi(z)\rangle = e^{-iHz}|\Psi(0)\rangle$, and the probability $P_j(z)=|c_j(z)|^2$ at output length $z$ identifies candidate minima. The key identity is that quantum tunneling allows the walker to pass through energy barriers, unlike classical gradient descent. The local annealing step prepares a low-energy input state by slowly reducing the potential in one well, and the varying Hamiltonian technique increases the potential depth along $z$ to sharpen localization on the global minimum.

Core claim. The central claim is that a photonic chip implementing quantum tunneling walk (QTW) can solve non-convex optimization problems by using the quantum tunneling effect to traverse barriers that trap classical methods. The authors experimentally verify that the average probability distribution of light exiting a waveguide array, whose refractive index encodes the objective function, highlights local minima in one- and two-dimensional landscapes. By adding a varying Hamiltonian (gradually increasing the potential depth along the propagation direction), they further show that the output intensity can concentrate on the global minimum in a landscape with uneven local minima. The paper thus asserts that QTS is a successful and efficient non-convex optimization solver that demonstrates quantum speedups over classical stochastic gradient descent.

The premise it rests on. The claim of quantum speedups over classical SGD assumes that the comparison of propagation length to SGD iterations is meaningful and fair, but the paper itself states that the axes are fundamentally different, the factor of 60 is only a display convention, and the data do not allow an end-to-end speedup estimate.

If the paper is right (Pith's inferences, not the authors')

If QTS works as claimed, photonic analog simulators could provide a complementary route to classical digital methods for non-convex optimization, especially for problems with high, thin barriers where tunneling is exponentially faster than thermal activation.

The demonstration that a fixed photonic chip can solve a specific landscape suggests that reconfigurable photonic platforms could, in principle, be programmed to tackle a family of optimization problems without repeated fabrication.

The local annealing preparation method offers a practical way to approximate the ground state of a potential well, which is required for the theoretical quantum speedup proven in Ref. [5].

The varying Hamiltonian technique gives a concrete mechanism to improve search accuracy and identify the global minimum, potentially extending the reach of waveguide-based optimization beyond simple local-minima detection.

The paper's claim of quantum speedups rests on a comparison to a simple stochastic gradient descent baseline, not to state-of-the-art classical optimizers like Adam or quasi-Newton methods. A stronger test would compare QTS to classical algorithms that also leverage non-local exploration, such as simulated annealing or evolutionary strategies, on the same landscape.

The experimental setup uses classical laser light, not single photons or entangled states, so the observed dynamics are classical wave optics obeying the same Schrödinger-like equation. True quantum advantage would require demonstrating that multi-photon interference (e.g., indistinguishable photon pairs) yields a computational benefit beyond what classical coherent light can achieve.

The scaling of the waveguide approach is severely limited: an $n$-dimensional grid with $k$ points per dimension requires $k^n$ waveguides. Without a more compact encoding, the method cannot address high-dimensional problems typical of machine learning. The authors acknowledge this but do not propose a concrete path forward beyond mentioning additive structure or low effective dimension as possibl

A critical open question is whether the local annealing step can be performed efficiently for an arbitrary landscape without prior knowledge of which well contains the low-energy state. The paper assumes that a rough local minimum is known, which may not be available in general non-convex problems.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

Non-convex optimization plays an important role in modern machine learning since plenty of real-world problems have objective functions or constraints with non-convex properties. The non-convex nature of landscapes gives rise to numerous local minima and saddle points, dramatically hindering the ability of gradient descent-based methods to explore the landscape and obtain global optima. Here we present quantum tunneling search (QTS), an algorithm based on quantum tunneling walk and local annealing, which effectively explores the non-convex landscape by leveraging the quantum tunneling effect. In experiments, we map an objective function to the potential term of the waveguide Hamiltonian by controlling the refractive index with the femtosecond direct laser writing technique. With a proper input state, the average probability distributions on 10-micrometer-spaced 29 x 1 and 21 x 9 waveguide arrays highlight the local minima of various 1D and 2D non-convex landscapes, respectively. By further introducing the varying Hamiltonian approach, more efficient search and even the identification of the global minimum are achieved. Our work experimentally verifies that QTS is a successful and efficient non-convex optimization solver, demonstrating quantum speedups over classical stochastic gradient descent (SGD), which serves as a practical application of near-term analog quantum simulators in the field of machine learning.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-13

The authors' main claim, as the classifier picked it out 1.00

“Our work experimentally verifies that QTS is a successful and efficient non-convex optimization solver, demonstrating quantum speedups over classical stochastic gradient descent (SGD), which serves as a practical application of near-term analog quantum simulators in the field of machine learning.”

How the main claim is stated: asserted 1.00

What the abstract itself states, as probabilities: claims first 0.04 · states limitation 0.10 · human participants 0.02 · controlled comparison 0.77 · randomized assignment 0.01 · states quantitative result 0.56 · compares with prior results 0.91 · mentions code or data availability 0.02

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
Non-convex optimization plays an important role in modern machine learning since plenty of real-world problems have objective functions or constraints with non-convex properties. background 1.00 0.04 0.02 asserted 1.00
The non-convex nature of landscapes gives rise to numerous local minima and saddle points, dramatically hindering the ability of gradient descent-based methods to explore the landscape and obtain global optima. background 0.92 0.05 0.02 asserted 1.00
Here we present quantum tunneling search (QTS), an algorithm based on quantum tunneling walk and local annealing, which effectively explores the non-convex landscape by leveraging the quantum tunneling effect. contribution 0.80 0.83 0.13 asserted 1.00
In experiments, we map an objective function to the potential term of the waveguide Hamiltonian by controlling the refractive index with the femtosecond direct laser writing technique. method 1.00 0.96 0.02 asserted 0.99
With a proper input state, the average probability distributions on 10-micrometer-spaced 29 x 1 and 21 x 9 waveguide arrays highlight the local minima of various 1D and 2D non-convex landscapes, respectively. result 0.90 0.96 0.04 asserted 1.00
By further introducing the varying Hamiltonian approach, more efficient search and even the identification of the global minimum are achieved. result 0.65 0.91 0.10 asserted 0.97
Our work experimentally verifies that QTS is a successful and efficient non-convex optimization solver, demonstrating quantum speedups over classical stochastic gradient descent (SGD), which serves as a practical application of near-term analog quantum simulators in the field of machine learning. interpretation 0.53 0.94 0.92 asserted 1.00

What kind of paper the classifier reports this is: method or tool 0.93 · evidence from observation or experiment 1.00

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-13: 1 request, 5,822 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

References (38)
  1. [5] Y. Liu, W. J. Su, and T. Li, On quantum speedups for nonconvex optimization via quantum tunneling walks, Quantum7, 1030 (2023) on Pith
  2. [1] P. Jain, P. Kar,et al., Non-convex optimization for ma- chine learning, Foundations and Trends®in Machine Learning10, 142 (2017) on Pith
  3. [2] I. Goodfellow, Y. Bengio, and A. Courville,Deep Learn- ing(MIT Press, 2016) on Pith
  4. [3] R. Meka, P. Jain, C. Caramanis, and I. S. Dhillon, Rank minimization via online learning, inProceedings of the 25th International Conference on Machine learning (2008) pp. 656–663 on Pith
  5. [4] X. Glorot and Y. Bengio, Understanding the difficulty of training deep feedforward neural networks, inProceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics(JMLR Workshop and Confer- ence Proceedings, 2010) pp. 249–256 on Pith
  6. [6] P. Rebentrost, M. Schuld, L. Wossnig, F. Petruccione, and S. Lloyd, Quantum gradient descent and newton’s method for constrained polynomial optimization, New Journal of Physics21, 073023 (2019) on Pith
  7. [7] J. Leng, E. Hickman, J. Li, and X. Wu, Quantum Hamil- tonian descent (2023), arXiv:2303.01471 on PithPith's reviewarXiv
  8. [8] Z. Chen, Y. Lu, H. Wang, Y. Liu, and T. Li, Quantum langevin dynamics for optimization, Communications in Mathematical Physics406, 52 (2025) on Pith
  9. [9] C.-F. Chen, H.-Y. Huang, J. Preskill, and L. Zhou, Lo- cal minima in quantum systems, Nature Physics21, 654 (2025) on Pith
  10. [10] Y. Wang, Y.-H. Lu, J. Gao, Y.-J. Chang, R.-J. Ren, Z.- Q. Jiao, Z.-Y. Zhang, and X.-M. Jin, Topologically pro- tected polarization quantum entanglement on a photonic chip, Chip1, 100003 (2022) on Pith
  11. [11] T. Kadowaki and H. Nishimori, Quantum annealing in the transverse ising model, Physical Review E , 5355 (1998) on Pith
  12. [12] S. Jiang, K. A. Britt, A. J. McCaskey, T. S. Humble, and S. Kais, Quantum annealing for prime factorization, Scientific Reports8, 17667 (2018) on Pith
  13. [13] D. Venturelli and A. Kondratyev, Reverse quantum an- 6 nealing approach to portfolio optimization problems, Quantum Machine Intelligence1, 17 (2019) on Pith
  14. [14] X.-J. Yuan, Z.-Q. Chen, Y.-D. Liu, Z. Xie, Y.-Z. Liu, X.-M. Jin, X. Wen, and H. Tang, Quantum support vec- tor machines for aerodynamic classification, Intelligent Computing2, 0057 (2023) on Pith
  15. [15] A. Mott, J. Job, J.-R. Vlimant, D. Lidar, and M. Spirop- ulu, Solving a higgs optimization problem with quantum annealing for machine learning, Nature550, 375 (2017) on Pith
  16. [16] A. Grabarits, F. Balducci, B. C. Sanders, and A. del Campo, Non-adiabatic quantum optimization for cross- ing quantum phase transitions, Physical Review A111, 012215 (2025) on Pith
  17. [17] S. V. Isakov, G. Mazzola, V. N. Smelyanskiy, Z. Jiang, S. Boixo, H. Neven, and M. Troyer, Understanding quan- tum tunneling through quantum monte carlo simulations, Physical Review Letters117, 180402 (2016) on Pith
  18. [18] Y. Zhan, Z. Ding, J. Huhn, J. Gray, J. Preskill, G. K.-L. Chan, and L. Lin, Rapid quantum ground state prepa- ration via dissipative dynamics, Physical Review X16, 011004 (2026) on Pith
  19. [19] H. Tang, L. Banchi, T.-Y. Wang, X.-W. Shang, X. Tan, W.-H. Zhou, Z. Feng, A. Pal, H. Li, C.-Q. Hu, M. S. Kim, and X.-M. Jin, Generating haar-uniform random- ness using stochastic quantum walks on a photonic chip, Physical Review Letters128, 050503 (2022) on Pith
  20. [20] M. Ohkuwa, H. Nishimori, and D. A. Lidar, Reverse an- nealing for the fully connected p-spin model, Physical Review A98, 022314 (2018) on Pith
show 18 more
  1. [21] H. Tang, X.-W. Shang, Z.-Y. Shi, T.-S. He, Z. Feng, T.-Y. Wang, R. Shi, H.-M. Wang, X. Tan, X.-Y. Xu,et al., Sim- ulating photosynthetic energy transport on a photonic network, npj Quantum Information10, 29 (2024) on Pith
  2. [22] Z. Lin, J. Li, W. Song, X. Li, H. Xin, X. Long, C. Chen, S. Zhu, T. Li, and S. Zhang, Artificial gauge fields for sculpting topological modes on photonic chips, Nature Communications17, 4829 (2026), published online 04 April 2026 on Pith
  3. [23] K. Kandasamy, J. Schneider, and B. Poczos, High di- mensional bayesian optimisation and bandits via additive models, inProceedings of the 32nd International Con- ference on Machine Learning, Proceedings of Machine Learning Research, Vol. 37 (PMLR, 2015) pp. 295–304 on Pith
  4. [24] Z. Wang, M. Zoghi, F. Hutter, D. Matheson, and N. de Freitas, Bayesian optimization in high dimensions via random embeddings, inProceedings of the Twenty- Third International Joint Conference on Artificial Intel- ligence(2013) pp. 1778–1784 on Pith
  5. [25] P. L. McMahon, A. Marandi, Y. Haribara, R. Hamerly, C. Langrock, S. Tamate, T. Inagaki, H. Takesue, S. Utsunomiya, K. Aihara, R. L. Byer, M. M. Fejer, H. Mabuchi, and Y. Yamamoto, A fully programmable 100-spin coherent ising machine with all-to-all connec- tions, Science354, 6... on Pith
  6. [26] Y. Wang, K. Liao, K. Zhang, Z. Du, Z. Wang, B. Ni, T. Xu, S. Feng, Y. Yang, Q.-F. Yang, Q. Sun, X. Hu, and Q. Gong, Reconfigurable versatile integrated pho- tonic computing chip, eLight5, 20 (2025) on Pith
  7. [27] Z. Cao, H.-J. Du, X.-J. Yuan,et al., Programmable three- dimensional photonic neural network chip, Nature Com- munications17, 5476 (2026) on Pith
  8. [28] Y. Zheng, H. Zhong, H. Zhang, L. Song, J. Liu, Y. Liang, Z. Liu, J. Chen, J. Zhou, Z. Fang, M. Wang, L. Li, R. Wu, and Y. Cheng, Electro-optically programmable photonic circuits enabled by wafer-scale integration on thin-film lithium niobate, Physical Review Research5, 033206 (2023) on Pith
  9. [29] Q. Wang, E. T. F. Rogers, B. Gholipour, C.-M. Wang, G. Yuan, J. Teng, and N. I. Zheludev, Optically reconfig- urable metasurfaces and photonic devices based on phase change materials, Nature Photonics10, 60 (2016) on Pith
  10. [30] D. P´ erez-L´ opez, A. L´ opez, P. DasMahapatra, and J. Cap- many, Multipurpose self-configuration of programmable photonic circuits, Nature Communications11, 6359 (2020) on Pith
  11. [31] B. Wu, W. Zhang, S. Zhang, H. Zhou, Z. Ruan, M. Li, D. Huang, J. Dong, and X. Zhang, A monolithically in- tegrated optical ising machine, Nature Communications 16, 4296 (2025) on Pith
  12. [32] W. Fu, X. Shi, S. S. Mohanraj, L. Shi, Y. Gao, Z. Wang, J. Wang, X. Chen, L. Qi, P. Aashna, G. Chen, D. Zhu, and A. Danner, Passive all-optical nonlinear neuron ac- tivation via PPLN nanophotonic waveguides, eLight6, 9 (2026) on Pith
  13. [33] Y. Yang, R. J. Chapman, B. Haylock, F. Lenzini, Y. N. Joglekar, M. Lobino, and A. Peruzzo, Programmable high-dimensional Hamiltonian in a photonic waveguide array, Nature Communications15, 50 (2024) on Pith
  14. [34] C. H. Bennett, E. Bernstein, G. Brassard, and U. Vazi- rani, Strengths and weaknesses of quantum computing, SIAM Journal on Computing26, 1510 (1997) on Pith
  15. [35] X.-L. Zhang, F. Yu, Z.-G. Chen, Z.-N. Tian, Q.-D. Chen, H.-B. Sun, and G. Ma, Non-abelian braiding on photonic chips, Nature Photonics16, 390 (2022) on Pith
  16. [36] Z. Hu, D. Bongiovanni, Z. Wang, X. Wang, D. Song, J. Xu, R. Morandotti, H. Buljan, and Z. Chen, Topolog- ical orbital angular momentum extraction and twofold protection of vortex transport, Nature Photonics , 1 (2024) on Pith
  17. [37] M. Ehrhardt, C. Dittel, M. Heinrich, and A. Szameit, Topological hong-ou-mandel interference, Science384, 1340 (2024) on Pith
  18. [38] X.-Y. Xu, X.-W. Wang, D.-Y. Chen, C. M. Smith, and X.-M. Jin, Quantum transport in fractal networks, Na- ture Photonics15, 703 (2021). 7 SUPPLEMENT AR Y MA TERIAL: “NON-CONVEX OPTIMIZA TION VIA QUANTUM TUNNELING SEARCH ON A PHOTONIC CHIP” Supplementary Note 1: Theoretical Anal... on Pith

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment