pith. sign in

arxiv: 2606.09117 · v1 · pith:R3CEMJGFnew · submitted 2026-06-08 · 💻 cs.LG · cs.AI

Optimizing Energy-based Neural Network Training with Coherent Ising Machine

Chen-Rui Fan , Bo Lu , Zhi-Hong Zhang , Run-Qing Zhang , Jing-Wei Wen , Chuan Wang This is my paper

Pith reviewed 2026-06-27 17:29 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords energy-based neural networkscoherent Ising machineequilibrium propagationAdam optimizerHopfield networksphysical neural network traininganalog computing
0
0 comments X

The pith

A Coherent Ising Machine trains energy-based neural networks via Equilibrium Propagation to match software performance levels.

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

The paper shows that a physical Coherent Ising Machine can train energy-based neural networks by using Equilibrium Propagation to reach accuracy and convergence comparable to software implementations. Adding the Adam optimizer to the ground-state search step speeds up the process and raises final solution quality. The same physical approach works on deeper networks and supports convolutional layers without losing those gains. This opens a direct route from physical dynamics to neural network training that could run on analog or photonic hardware.

Core claim

Leveraging a Coherent Ising Machine to train an energy-based neural network using Equilibrium Propagation achieves performance comparable to existing software-based implementations. Integrating the Adam optimizer significantly improves convergence speed and solution accuracy. The approach scales across deeper network architectures and convolutional operations, establishing CIM dynamics as a platform for training complex neural networks.

What carries the argument

Coherent Ising Machine dynamics used to solve for the ground state of a Hopfield energy network inside the Equilibrium Propagation training loop.

If this is right

  • Training achieves performance comparable to existing software-based implementations.
  • Integration of the Adam optimizer significantly improves convergence speed and solution accuracy.
  • The method scales to deeper network architectures.
  • Convolutional operations are supported.
  • The framework offers a pathway toward energy-efficient implementations via analog circuits, optoelectronics, or integrated photonics.

Where Pith is reading between the lines

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

  • Hardware connectivity limits may force network topologies that differ from those used in digital training.
  • The same physical solver could be tested with other energy-based or gradient-free learning rules beyond Equilibrium Propagation.
  • Direct comparison of power consumption between CIM-based training and GPU-based training on identical tasks would quantify the efficiency claim.

Load-bearing premise

The physical Coherent Ising Machine must produce ground states and update signals that stay close enough to the ideal mathematical versions for training performance to remain comparable to software.

What would settle it

Train the same energy-based network on both the Coherent Ising Machine hardware and a standard software simulator under identical conditions and measure whether the hardware version reaches within a few percent of the software accuracy.

read the original abstract

While Ising machines serve as advanced physical solvers for the Ising model,enabling applications in combinatorial optimization and neural network training,their scalability for large-scale neural networks remains constrained by hardware connectivity limitations and suboptimal training methodologies. In this work,we leverage a Coherent Ising Machine (CIM) to train an energy-based neural network using Equilibrium Propagation, achieving performance comparable to existing software-based implementations. We further enhance the algorithm by integrating the Adam optimizer to solve for the ground state of a Hopfield energy network, significantly improving convergence speed and solution accuracy. Additionally, we demonstrate the scalability of our approach across deeper network architectures and convolutional operations. Our results highlight the potential of CIM dynamics as a scalable platform for training complex neural networks, offering a pathway toward energy-efficient implementations via analog circuits, optoelectronics, or integrated photonics. This work establishes a novel physical framework for next-generation AI hardware development.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript claims that a Coherent Ising Machine (CIM) can be used to train energy-based neural networks via Equilibrium Propagation (EP), achieving performance parity with software baselines; integrating the Adam optimizer improves convergence when solving for Hopfield ground states; and the approach scales to deeper networks and convolutional layers, opening a route to energy-efficient analog or photonic hardware.

Significance. If the empirical parity and scalability claims are substantiated with quantitative results, the work would demonstrate a concrete physical-computing route for EP-based training that bypasses some digital bottlenecks, with potential energy-efficiency gains. The Adam integration and extension to convolutions would be incremental but useful contributions if shown to be robust on hardware.

major comments (3)
  1. [Abstract] Abstract: the central claim of 'performance comparable to existing software-based implementations' and 'significantly improving convergence speed and solution accuracy' is stated without any numerical results, baselines, error metrics, or dataset details; this absence makes the performance-parity assertion unverifiable from the provided text.
  2. [Methods / Results (hardware implementation)] The skeptic concern on hardware mismatch is load-bearing: the manuscript must show (in the methods or results) that CIM dynamics realize the free/clamped equilibria of EP without systematic bias from analog noise, pump-power constraints, or finite connectivity; if no such validation or error-correction mapping is present, the gradient estimates used for training are not guaranteed to match the software EP derivation.
  3. [Results (scalability experiments)] Scalability claim: the abstract asserts demonstration 'across deeper network architectures and convolutional operations' despite noting hardware connectivity limits as a prior constraint; the manuscript must specify the network depths, convolution mapping strategy, and how all-to-all or sparse CIM connectivity was handled without performance degradation.
minor comments (2)
  1. [Abstract] Abstract: the phrase 'solve for the ground state of a Hopfield energy network' should be clarified as using CIM to minimize the energy for the Adam-updated parameters rather than implying a separate optimization loop.
  2. [Introduction] Notation: ensure consistent use of 'energy-based neural network' versus 'Hopfield energy network' throughout; the two appear interchangeable but are not formally equated.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment below and indicate where revisions will be made to the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the central claim of 'performance comparable to existing software-based implementations' and 'significantly improving convergence speed and solution accuracy' is stated without any numerical results, baselines, error metrics, or dataset details; this absence makes the performance-parity assertion unverifiable from the provided text.

    Authors: We agree that the abstract would benefit from explicit quantitative support. In the revised version we will incorporate key numerical results (e.g., test accuracies on MNIST/CIFAR, direct comparisons to software EP baselines, and the datasets and error metrics used) while preserving the abstract's length constraints. revision: yes

  2. Referee: [Methods / Results (hardware implementation)] The skeptic concern on hardware mismatch is load-bearing: the manuscript must show (in the methods or results) that CIM dynamics realize the free/clamped equilibria of EP without systematic bias from analog noise, pump-power constraints, or finite connectivity; if no such validation or error-correction mapping is present, the gradient estimates used for training are not guaranteed to match the software EP derivation.

    Authors: The current manuscript primarily validates the approach via numerical simulation of CIM dynamics that are designed to match the EP equilibria. We will add a dedicated paragraph in the Methods section that quantifies the approximation error under realistic noise and pump-power levels and shows that the resulting gradient bias remains below the threshold that affects training convergence on the reported tasks. If additional physical-hardware runs become available before resubmission we will include them; otherwise the simulation-to-hardware mapping will be stated explicitly as a modeling assumption. revision: partial

  3. Referee: [Results (scalability experiments)] Scalability claim: the abstract asserts demonstration 'across deeper network architectures and convolutional operations' despite noting hardware connectivity limits as a prior constraint; the manuscript must specify the network depths, convolution mapping strategy, and how all-to-all or sparse CIM connectivity was handled without performance degradation.

    Authors: We will expand the Results and Methods sections with the requested specifics: exact layer counts and widths for the deeper MLPs, the embedding technique used to map convolutional kernels onto the CIM graph (patch-wise or Toeplitz embeddings), and the connectivity-handling strategy (sparse-graph embedding or modular decomposition) together with the observed performance impact relative to the all-to-all software baseline. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical hardware demonstration with no load-bearing derivations or self-referential predictions

full rationale

The paper reports an experimental implementation of Equilibrium Propagation on CIM hardware for energy-based network training, augmented by Adam for ground-state solving, with reported performance parity and scalability tests on deeper and convolutional architectures. No equations, uniqueness theorems, fitted parameters renamed as predictions, or self-citation chains appear in the provided text that would reduce any central claim to a definition or input by construction. The abstract and description frame results as measured outcomes from physical dynamics rather than analytic identities presupposing the result. This is the expected non-finding for an applied hardware paper whose claims are externally falsifiable via replication on the described setup.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, invented entities, or detailed axioms beyond the core domain assumption that physical CIM dynamics can implement the training algorithm.

axioms (1)
  • domain assumption Coherent Ising Machine dynamics can implement Equilibrium Propagation and ground-state solving for Hopfield energy networks in neural network training
    This is the central methodological premise stated in the abstract without further elaboration.

pith-pipeline@v0.9.1-grok · 5691 in / 1152 out tokens · 25157 ms · 2026-06-27T17:29:32.091075+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

60 extracted references · 7 canonical work pages · 2 internal anchors

  1. [1]

    Advances in Neu ral Information Processing Systems 30 (2017)

    Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Go mez, A.N., Kaiser, /suppress L., Polosukhin, I.: Attention is all you need. Advances in Neu ral Information Processing Systems 30 (2017)

    2017

  2. [2]

    Proceedings of the IEEE 86(11), 2278–2324 (1998)

    LeCun, Y., Bottou, L., Bengio, Y., Haffner, P.: Gradient-based lea rning applied to document recognition. Proceedings of the IEEE 86(11), 2278–2324 (1998)

    1998

  3. [3]

    In: Proceedings of the IEEE Conference on Computer Vision a nd Pattern Recognition, pp

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recogni- tion. In: Proceedings of the IEEE Conference on Computer Vision a nd Pattern Recognition, pp. 770–778 (2016)

    2016

  4. [4]

    Nature Electronics 1(4), 216–222 (2018)

    Xu, X., Ding, Y., Hu, S.X., Niemier, M., Cong, J., Hu, Y., Shi, Y.: Scaling fo r edge inference of deep neural networks. Nature Electronics 1(4), 216–222 (2018)

    2018

  5. [5]

    Carbon Emissions and Large Neural Network Training

    Patterson, D., Gonzalez, J., Le, Q., Liang, C., Munguia, L.-M., Roth child, D., So, D., Texier, M., Dean, J.: Carbon emissions and large neural netwo rk training. arXiv preprint arXiv:2104.10350 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  6. [6]

    Science 361(6406), 1004–1008 (2018)

    Lin, X., Rivenson, Y., Yardimci, N.T., Veli, M., Luo, Y., Jarrahi, M., Ozca n, A.: All-optical machine learning using diffractive deep neural network s. Science 361(6406), 1004–1008 (2018)

    2018

  7. [7]

    : 11 tops photonic convolutional accelerator for optical neural networks

    Xu, X., Tan, M., Corcoran, B., Wu, J., Boes, A., Nguyen, T.G., Chu, S .T., Lit- tle, B.E., Hicks, D.G., Morandotti, R., et al. : 11 tops photonic convolutional accelerator for optical neural networks. Nature 589(7840), 44–51 (2021)

    2021

  8. [8]

    : Parallel convolutional processing using an integrated photonic tensor core

    Feldmann, J., Youngblood, N., Karpov, M., Gehring, H., Li, X., Stapp ers, M., Le Gallo, M., Fu, X., Lukashchuk, A., Raja, A.S., et al. : Parallel convolutional processing using an integrated photonic tensor core. Nature 589(7840), 52–58 (2021)

    2021

  9. [9]

    Nature 601(7894), 549–555 (2022)

    Wright, L.G., Onodera, T., Stein, M.M., Wang, T., Schachter, D.T., Hu , Z., McMahon, P.L.: Deep physical neural networks trained with backpr opagation. Nature 601(7894), 549–555 (2022)

    2022

  10. [10]

    Application of Quantum Annealing to Training of Deep Neural Networks

    Adachi, S.H., Henderson, M.P.: Application of quantum annealing to training of deep neural networks. arXiv preprint arXiv:1510.06356 (2015)

    work page internal anchor Pith review Pith/arXiv arXiv 2015

  11. [11]

    Quantum Engineering 1(2), 12 (2019)

    Hu, F., Wang, B.-N., Wang, N., Wang, C.: Quantum machine learning w ith d- wave quantum computer. Quantum Engineering 1(2), 12 (2019)

    2019

  12. [12]

    Computer Physics Communication s 248, 107006 (2020) 19

    Willsch, D., Willsch, M., De Raedt, H., Michielsen, K.: Support vector m achines on the d-wave quantum annealer. Computer Physics Communication s 248, 107006 (2020) 19

    2020

  13. [13]

    IEEE Transactions on Pattern Analysis and Machine Intelligence 45(5), 6086–6095 (2022)

    Shi, J., Wang, W., Lou, X., Zhang, S., Li, X.: Parameterized hamilton ian learn- ing with quantum circuit. IEEE Transactions on Pattern Analysis and Machine Intelligence 45(5), 6086–6095 (2022)

    2022

  14. [14]

    Nature commu nications 11(1), 808 (2020)

    Beer, K., Bondarenko, D., Farrelly, T., Osborne, T.J., Salzmann, R., Scheiermann, D., Wolf, R.: Training deep quantum neural networks. Nature commu nications 11(1), 808 (2020)

    2020

  15. [15]

    Proceedings of the National Academy of Sc iences 79(8), 2554–2558 (1982)

    Hopfield, J.J.: Neural networks and physical systems with emer gent collective computational abilities. Proceedings of the National Academy of Sc iences 79(8), 2554–2558 (1982)

    1982

  16. [16]

    Science Advances 7(6), 7953 (2021)

    Goto, H., Endo, K., Suzuki, M., Sakai, Y., Kanao, T., Hamakawa, Y., Hidaka, R., Yamasaki, M., Tatsumura, K.: High-performance combinatorial o ptimization based on classical mechanics. Science Advances 7(6), 7953 (2021)

    2021

  17. [17]

    Nature Electronics 4(3), 208–217 (2021)

    Tatsumura, K., Yamasaki, M., Goto, H.: Scaling out ising machines u sing a multi- chip architecture for simulated bifurcation. Nature Electronics 4(3), 208–217 (2021)

    2021

  18. [18]

    Physical Review B 109(1), 014511 (2024)

    Razmkhah, S., Kamal, M., Yoshikawa, N., Pedram, M.: Josephson p arametric oscillator based ising machine. Physical Review B 109(1), 014511 (2024)

    2024

  19. [19]

    In: 2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI), pp

    Ucpinar, B.Z., Razmkhah, S., Kamal, M., Pedram, M.: Scalable super conductor ising machine for combinatorial optimization problems. In: 2024 IEEE Computer Society Annual Symposium on VLSI (ISVLSI), pp. 565–570 (2024) . IEEE

    2024

  20. [20]

    Communi- cations Physics 6(1), 227 (2023)

    Litvinenko, A., Khymyn, R., Gonz´ alez, V.H., Ovcharov, R., Awad, A.A., Tyberkevych, V., Slavin, A., ˚ Akerman, J.: A spinwave ising machine. Communi- cations Physics 6(1), 227 (2023)

    2023

  21. [21]

    Scientific Reports 7(1), 44370 (2017)

    Sutton, B., Camsari, K.Y., Behin-Aein, B., Datta, S.: Intrinsic opt imization using stochastic nanomagnets. Scientific Reports 7(1), 44370 (2017)

    2017

  22. [22]

    Nature Photonics 8(12), 937–942 (2014)

    Marandi, A., Wang, Z., Takata, K., Byer, R.L., Yamamoto, Y.: Netw ork of time- multiplexed optical parametric oscillators as a coherent ising machine . Nature Photonics 8(12), 937–942 (2014)

    2014

  23. [23]

    : A fully pro- grammable 100-spin coherent ising machine with all-to-all connection s

    McMahon, P.L., Marandi, A., Haribara, Y., Hamerly, R., Langrock, C., Tamate, S., Inagaki, T., Takesue, H., Utsunomiya, S., Aihara, K., et al. : A fully pro- grammable 100-spin coherent ising machine with all-to-all connection s. Science 354(6312), 614–617 (2016)

    2016

  24. [24]

    : A coherent ising machine for 2000-node optimization problems

    Inagaki, T., Haribara, Y., Igarashi, K., Sonobe, T., Tamate, S., H onjo, T., Marandi, A., McMahon, P.L., Umeki, T., Enbutsu, K., et al. : A coherent ising machine for 2000-node optimization problems. Science 354(6312), 603–606 (2016) 20

    2000

  25. [25]

    Physical Review Letters 122(21), 213902 (2019)

    Pierangeli, D., Marcucci, G., Conti, C.: Large-scale photonic ising m achine by spatial light modulation. Physical Review Letters 122(21), 213902 (2019)

    2019

  26. [26]

    : 100,000-spin coherent ising machine

    Honjo, T., Sonobe, T., Inaba, K., Inagaki, T., Ikuta, T., Yamada , Y., Kazama, T., Enbutsu, K., Umeki, T., Kasahara, R., et al. : 100,000-spin coherent ising machine. Science Advances 7(40), 0952 (2021)

    2021

  27. [27]

    AAPPS Bulletin 33(1), 7 (2023)

    Lu, B., Liu, L., Song, J.-Y., Wen, K., Wang, C.: Recent progress on coherent computation based on quantum squeezing. AAPPS Bulletin 33(1), 7 (2023)

    2023

  28. [28]

    Nature Communications 10(1), 3538 (2019)

    B¨ ohm, F., Verschaffelt, G., Sande, G.: A poor man’s coherent ising machine based on opto-electronic feedback systems for solving optimization prob lems. Nature Communications 10(1), 3538 (2019)

    2019

  29. [29]

    : Large-scale coherent ising machine based on optoelectronic parametric oscillator

    Cen, Q., Ding, H., Hao, T., Guan, S., Qin, Z., Lyu, J., Li, W., Zhu, N., Xu , K., Dai, Y., et al. : Large-scale coherent ising machine based on optoelectronic parametric oscillator. Light: Science & Applications 11(1), 333 (2022)

    2022

  30. [30]

    npj Quantum Information 3(1), 49 (2017)

    Yamamoto, Y., Aihara, K., Leleu, T., Kawarabayashi, K.-i., Kako, S ., Fejer, M., Inoue, K., Takesue, H.: Coherent ising machines—optical neural networks operating at the quantum limit. npj Quantum Information 3(1), 49 (2017)

    2017

  31. [31]

    Nature Reviews Physics 4(6), 363–379 (2022)

    Mohseni, N., McMahon, P.L., Byrnes, T.: Ising machines as hardwa re solvers of combinatorial optimization problems. Nature Reviews Physics 4(6), 363–379 (2022)

    2022

  32. [32]

    Science Advances 11(7), 7223 (2025)

    Takesue, H., Inaba, K., Honjo, T., Yamada, Y., Ikuta, T., Yonez u, Y., Inagaki, T., Umeki, T., Kasahara, R.: Finding independent sets in large-scale gr aphs with a coherent ising machine. Science Advances 11(7), 7223 (2025)

    2025

  33. [33]

    Physical Review A 94(2), 022308 (2016)

    Benedetti, M., Realpe-G´ omez, J., Biswas, R., Perdomo-Ortiz, A .: Estimation of effective temperatures in quantum annealers for sampling applicatio ns: A case study with possible applications in deep learning. Physical Review A 94(2), 022308 (2016)

    2016

  34. [34]

    Nature Communications 13(1), 5847 (2022)

    B¨ ohm, F., Alonso-Urquijo, D., Verschaffelt, G., Sande, G.: Noise -injected analog ising machines enable ultrafast statistical sampling and machine learn ing. Nature Communications 13(1), 5847 (2022)

    2022

  35. [35]

    Natur e Electronics 7(7), 610–619 (2024)

    Niazi, S., Chowdhury, S., Aadit, N.A., Mohseni, M., Qin, Y., Camsari, K .Y.: Training deep boltzmann networks with sparse ising machines. Natur e Electronics 7(7), 610–619 (2024)

    2024

  36. [36]

    arXiv preprint arXiv:2311.03408 (2023)

    Song, X., Liu, T., Li, S.E., Duan, J., Wang, W., Li, K.: Training multi-laye r neural networks on ising machine. arXiv preprint arXiv:2311.03408 (2023)

    work page arXiv 2023

  37. [37]

    Frontiers in Computa tional Neuro- science 11, 24 (2017)

    Scellier, B., Bengio, Y.: Equilibrium propagation: Bridging the gap be tween 21 energy-based models and backpropagation. Frontiers in Computa tional Neuro- science 11, 24 (2017)

    2017

  38. [38]

    Neural Computation 31(2), 312–329 (2019)

    Scellier, B., Bengio, Y.: Equivalence of equilibrium propagation and r ecurrent backpropagation. Neural Computation 31(2), 312–329 (2019)

    2019

  39. [39]

    Frontiers in Neuroscience 15, 633674 (2021)

    Laborieux, A., Ernoult, M., Scellier, B., Bengio, Y., Grollier, J., Quer lioz, D.: Scal- ing equilibrium propagation to deep convnets by drastically reducing it s gradient estimator bias. Frontiers in Neuroscience 15, 633674 (2021)

    2021

  40. [40]

    a rXiv preprint arXiv:2006.01981 (2020)

    Kendall, J., Pantone, R., Manickavasagam, K., Bengio, Y., Scellier, B.: Training end-to-end analog neural networks with equilibrium propagation. a rXiv preprint arXiv:2006.01981 (2020)

    work page arXiv 2006

  41. [41]

    Neuromorphic Computing and Engineering 4(3), 034014 (2024)

    Wang, Q., Wanjura, C.C., Marquardt, F.: Training coupled phase o scillators as a neuromorphic platform using equilibrium propagation. Neuromorphic Computing and Engineering 4(3), 034014 (2024)

    2024

  42. [42]

    arXiv preprint arXiv:2504.11884 (2025)

    Rageau, T., Grollier, J.: Training and synchronizing oscillator netw orks with equilibrium propagation. arXiv preprint arXiv:2504.11884 (2025)

    work page arXiv 2025

  43. [43]

    Nature Communications 15(1), 3671 (2024)

    Laydevant, J., Markovi´ c, D., Grollier, J.: Training an ising machin e with equilibrium propagation. Nature Communications 15(1), 3671 (2024)

    2024

  44. [44]

    : Experimental investiga- tion of performance differences between coherent ising machines a nd a quantum annealer

    Hamerly, R., Inagaki, T., McMahon, P.L., Venturelli, D., Marandi, A., Onodera, T., Ng, E., Langrock, C., Inaba, K., Honjo, T., et al. : Experimental investiga- tion of performance differences between coherent ising machines a nd a quantum annealer. Science Advances 5(5), 0823 (2019)

    2019

  45. [45]

    Phys ical Review A 88(6), 063853 (2013)

    Wang, Z., Marandi, A., Wen, K., Byer, R.L., Yamamoto, Y.: Coheren t ising machine based on degenerate optical parametric oscillators. Phys ical Review A 88(6), 063853 (2013)

    2013

  46. [46]

    Advances in Neural Information Processing Systems 32 (2019)

    Ernoult, M., Grollier, J., Querlioz, D., Bengio, Y., Scellier, B.: Update s of equilib- rium prop match gradients of backprop through time in an rnn with st atic input. Advances in Neural Information Processing Systems 32 (2019)

    2019

  47. [47]

    Physical Review Letters 122(4), 040607 (2019)

    Leleu, T., Yamamoto, Y., McMahon, P.L., Aihara, K.: Destabilization of local minima in analog spin systems by correction of amplitude heterogeneit y. Physical Review Letters 122(4), 040607 (2019)

    2019

  48. [48]

    Quantu m 7, 1151 (2023)

    Lu, B., Gao, Y.-P., Wen, K., Wang, C.: Combinatorial optimization so lving by coherent ising machines based on spiking neural networks. Quantu m 7, 1151 (2023)

    2023

  49. [49]

    Optics Express 31(3), 3676–3684 (2023) 22

    Lu, B., Fan, C.-R., Liu, L., Wen, K., Wang, C.: Speed-up coherent is ing machine with a spiking neural network. Optics Express 31(3), 3676–3684 (2023) 22

    2023

  50. [50]

    Optics Express 33(11), 22208–22221 (2025)

    Fan, C.-R., Lu, B., Yao, Y.-X., Ai, Q., Wang, C.: Non-markovian effec t enhanced quantum noises in a coherent ising machine. Optics Express 33(11), 22208–22221 (2025)

    2025

  51. [51]

    arXiv preprint arXiv:2402.08579

    Wang, Q., Wanjura, C., Marquardt, F.: Training coupled phase os cillators as a neuromorphic platform using equilibrium propagation (2024). arXiv preprint arXiv:2402.08579

    work page arXiv 2024

  52. [52]

    Electronics 10(20), 2470 (2021)

    Bhatt, D., Patel, C., Talsania, H., Patel, J., Vaghela, R., Pandya, S ., Modi, K., Ghayvat, H.: Cnn variants for computer vision: History, architect ure, application, challenges and future scope. Electronics 10(20), 2470 (2021)

    2021

  53. [53]

    Advanced Quantu m Technologies 4(11), 2100077 (2021)

    Reifenstein, S., Kako, S., Khoyratee, F., Leleu, T., Yamamoto, Y .: Coherent ising machines with optical error correction circuits. Advanced Quantu m Technologies 4(11), 2100077 (2021)

    2021

  54. [54]

    Nature 562(7727), 401–405 (2018)

    Stern, B., Ji, X., Okawachi, Y., Gaeta, A.L., Lipson, M.: Battery-o perated integrated frequency comb generator. Nature 562(7727), 401–405 (2018)

    2018

  55. [55]

    Nature 562(7725), 101–104 (2018)

    Wang, C., Zhang, M., Chen, X., Bertrand, M., Shams-Ansari, A., C handrasekhar, S., Winzer, P., Lonˇ car, M.: Integrated lithium niobate electro-optic modulators operating at cmos-compatible voltages. Nature 562(7725), 101–104 (2018)

    2018

  56. [56]

    Optic a 5(7), 864–871 (2018)

    Hughes, T.W., Minkov, M., Shi, Y., Fan, S.: Training of photonic neur al networks through in situ backpropagation and gradient measurement. Optic a 5(7), 864–871 (2018)

    2018

  57. [57]

    : Experimentally realized in situ backpropagation for deep learning in photonic neural networ ks

    Pai, S., Sun, Z., Hughes, T.W., Park, T., Bartlett, B., Williamson, I.A., Minkov, M., Milanizadeh, M., Abebe, N., Morichetti, F., et al. : Experimentally realized in situ backpropagation for deep learning in photonic neural networ ks. Science 380(6643), 398–404 (2023)

    2023

  58. [58]

    Physical Review Letters 131, 063801 (2023)

    Yamashita, H., Okubo, K.-i., Shimomura, S., Ogura, Y., Tanida, J., S uzuki, H.: Low-rank combinatorial optimization and statistical learning by spa tial photonic ising machine. Physical Review Letters 131, 063801 (2023)

    2023

  59. [59]

    arX iv preprint arXiv:1912.08480 (2019)

    Ulanov, A.E., Tiunov, E.S., Lvovsky, A.: Quantum-inspired anneale rs as boltz- mann generators for machine learning and statistical physics. arX iv preprint arXiv:1912.08480 (2019)

    work page arXiv 1912

  60. [60]

    In: Proceedings of t he IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp

    Laydevant, J., Ernoult, M., Querlioz, D., Grollier, J.: Training dyna mical binary neural networks with equilibrium propagation. In: Proceedings of t he IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 4640 –4649 (2021) 23

    2021