Pith. sign in

REVIEW 4 major objections 5 minor 33 references

Annealing Machine-assisted Learning of Graph Neural Network for Combinatorial Optimization

T0 review · 4 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read The paper claims that an annealing machine can solve combinatorial optimization problems larger than its native variable limit by supervising local graph neural networks on Louvain-compressed subproblems and using the learned features to…

desk verdict A plausible AM-GNN transfer pipeline, but the AM's specific contribution is untested because the only control misses the value of any external supervision. read the letter →

arxiv 2501.05845 v1 pith:MJW2377Y submitted 2025-01-10 cs.AI cs.LG

classification cs.AIcs.LG
keywords combinatorialoptimizationannealingmachinegraphneuralnetworkQUBOLouvaindecompositionmultiresolutionguidancemaximumindependentsetknowledgetransfer
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

Annealing machines are precise solvers for QUBO problems, but each machine has a hard cap on the number of variables it can handle. This paper proposes to break that cap by compressing the problem graph with Louvain community detection, letting the annealing machine solve each compressed graph, and using those solutions as labels for small local graph neural networks. The node representations those networks learn are mapped back to the original graph and averaged to initialize a larger GNN, which then solves the original problem unsupervised. The authors report that on maximum independent set instances with 100k and 150k nodes, this AM-guided pipeline lowers the final Hamiltonian loss by roughly 25–35% compared with the same GNN architecture without AM guidance, with fewer constraint violations. If the transfer works as claimed, it offers a practical way to apply a fixed annealing machine to problems far beyond its hardware limit.

What carries the argument

The load-bearing mechanism is the AM-guided multiresolution GNN pipeline. It consists of three coupled pieces: (i) Louvain decomposition produces a sequence of compressed graphs $\{G_i\}_{i=1}^s$ with a mapping from each artificial node back to a community of original nodes; (ii) a Guiding block trains a local GNN on each $G_i$ with a loss that combines the QUBO Hamiltonian $H_{QUBO}$ and the squared-error alignment $\ell(x^{GNN}_{i,t}, x^{AM}_i) = \|x^{GNN}_{i,t} - x^{AM}_i\|^2$ against the annealing machine's binary solution, so the AM acts as a teaching signal; (iii) a Mapping module projects each learned node vector $\bar{f}_{n_i}$ onto the original nodes in that community, weighted by each node's degree divided by the community's total degree, then averages the $s$ projected feature matrices $\bar{R}_i$ to form the initial features $R$ for the main GNN solver on the original graph $G_P$, which is trained by minimizing $H_{QUBO}$ alone. The key design choice is that the AM never sees the original large graph; its knowledge enters only through the features that initialize the global solver.

What would settle it

Take graphs small enough that exact or high-quality solutions are available (e.g., random d-regular graphs up to a few thousand nodes), run the full pipeline, and check two things: whether the AM's solution on the compressed graph, lifted back through the Louvain mapping, correlates with the true optimum on the original graph, and whether mrGNN+AM's lower Hamiltonian loss corresponds to actually better objective values (e.g., larger independent sets or more balanced cuts). If the lifted AM solutions are uncorrelated with optimality while the pipeline still shows lower Hamiltonian loss, the central claim of knowledge transfer fails and the apparent gain is an artifact of measuring quality with the same Hamiltonian the solver minimizes.

Watch

Extended reading notes

Core claim

The central discovery is that the annealing machine never needs to see the original large graph: its solutions on Louvain-compressed graphs act as supervision for local GNNs, and the feature vectors learned by those local GNNs, lifted to original nodes through an inverse-degree-weighted Louvain mapping and averaged over resolutions, form an initialization for an unsupervised global GNN solver that is better than random initialization. On maximum independent set instances with 100k and 150k nodes, this AM-guided solver achieves 25–35% lower final Hamiltonian loss than the multiresolution GNN without AM guidance, with fewer violations; on MaxCut and Graph Partition the improvements are smaller and mostly appear at 150k nodes. The paper also reports that the AM-guided solver reduces late-training assignment shifts, and that classical heuristics time out on the larger graphs while the proposed pipeline finishes in about twelve minutes on average. The authors read these results as evidence that GNN flexibility can carry the annealing machine's accuracy to problem sizes beyond its native variable limit.

Load-bearing premise

The load-bearing premise is that the QUBO matrices built from Louvain-compressed graphs are aligned enough with the original problem's Hamiltonian that the annealing machine's solutions on those small graphs teach the local GNNs something true about the large graph; if that alignment fails, the transferred features are no better than random and the reported quality gains—measured by Hamiltonian loss without ground truth—would be an artifact.

Editorial extensions

If this is right

  • An annealing machine with a fixed variable limit can be applied to arbitrarily larger instances by letting a GNN carry its solution knowledge across resolutions, so hardware scaling is no longer the only route to larger problems.
  • The final solver runs on the original graph, so the output size is unbounded by the AM's limit; the AM only needs to handle the largest compressed subproblem.
  • On maximum independent set instances of 100k–150k nodes, the AM-guided solver lowers the final Hamiltonian loss by roughly 25–35% relative to the unsupervised multiresolution GNN, with fewer constraint violations.
  • AM guidance stabilizes training: the main solver exhibits fewer late-stage assignment shifts, which makes early stopping more reliable.
  • The pipeline remains practical in wall-clock terms: it finishes 150k-node MIS instances in about 12 minutes on average, whereas the tested classical heuristics time out after two hours.

Reading between the lines

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

  • The same distillation could apply to any expensive solver—quantum annealers, coherent Ising machines, or specialized CMOS annealers—so solver improvements become portable across hardware generations.
  • Because local GNNs are roughly 9× faster than the AM, one could run the AM once on a training set of compressed graphs, then deploy a purely GNN pipeline at inference time, trading a small quality loss for large speedups.
  • The inverse-degree weighting in the mapping module is a free design choice; ablating it (e.g., uniform averaging within communities) would reveal whether the reported gains come from the weighting or simply from community-aware feature pooling.
  • Testing on graphs with known optima would separate the paper's Hamiltonian-loss measure from true objective quality, since a lower Hamiltonian could in principle come from penalty-term exploitation rather than better combinatorial solutions.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. This paper proposes a pipeline that combines annealing-machine (AM) solvers with graph neural networks (GNNs) for large combinatorial optimization. Given an original graph G_P, the method applies Louvain compression to obtain a hierarchy of smaller graphs, solves each compressed graph on a CMOS annealing machine, uses the resulting binary solutions as supervised targets for local GNNs, aggregates the learned node representations into initial features for a global GNN solver that minimizes the QUBO Hamiltonian on the original graph. Experiments on synthetic d-regular graphs (n=50k, 100k, 150k; d=3,4,5) for MaxCut, MIS, and Graph Partition compare three variants (rGNN, mrGNN, mrGNN+AM), reporting Hamiltonian loss, violation counts, and wall-clock time, plus comparisons with greedy, HB, and a neural baseline on MIS. The main reported findings are relative loss improvements of roughly 25-35% for MIS at 100k and 150k nodes, smaller gains for MaxCut and GP, and mixed timing results.

Significance. If the improvements are attributable to the AM's guidance, the paper would demonstrate a practical scheme for extending AM-style solution quality to graphs beyond hardware limits, with a credible scale of 150k nodes and multiple problem formulations. The paper's strengths include evaluation on large synthetic graphs, reporting of constraint-violation counts alongside Hamiltonian loss, explicit wall-clock measurements, and comparisons with classical baselines. However, because the experimental design does not isolate the AM-specific contribution and lacks statistical error bars, the central causal claim is not yet established; the correct scope of the contribution depends on the additional controls recommended below.

major comments (4)
  1. [Section 3 (Multiresolution Guidance I and II); Table 4] The central claim that AM knowledge causes the gains is not tested. The comparison mrGNN vs mrGNN+AM conflates three factors: the Louvain hierarchy, the presence of any external supervision signal for the local GNNs, and the specific AM solutions. The paper explicitly assumes alignment between compressed QUBOs and the original problem ("there should be certain alignment," Section 3), but no experiment validates that assumption. A negative control is required: run the identical pipeline with random binary labels, and with cheap heuristic labels (e.g., greedy solutions on each compressed graph), in place of xAM_i. If these controls yield similar relative loss differences on the same test graphs, then the AM-specific "beyond limits" claim is not supported; if they do not, the control would establish causal attribution. This is the load-bearing issue for the paper's title and abstract claim.
  2. [Section 5 (Solution Quality and Convergence); Table 4] All relative loss differences are reported as single point estimates. The text says solutions were sampled 50 times per target graph, yet Table 4 gives no standard deviation, confidence interval, or significance test for Δrel. For the headline MIS numbers (25-35% at n=100k and 150k), the spread across degrees within a fixed n (e.g., 25.06%, 28.11%, 33.25% at 100k) is comparable to possible run-to-run noise. Reporting mean ± std over the 50 samples, and ideally paired differences between mrGNN and mrGNN+AM on identical graph instances, is necessary to establish that the improvements are not noise.
  3. [Abstract, Section 1, and Table 4] The abstract and introduction state that the approach reduces "the overall converge time" / "overall convergence time." Table 4's ΔT is total execution time with ΔT>0 defined as mrGNN being faster. Under that definition, mrGNN+AM is slower by 21-89% for MIS at n=100k and 150k, and is meaningfully faster mostly at n=50k. For MaxCut and GP, the timing differences are mixed and often small. Thus the general speed claim is contradicted by the reported data; the paper should either restrict the claim to the cases where it holds or report a separate convergence-time metric (e.g., epochs to convergence) if that is what is meant.
  4. [Section 5 (Solution Quality and Convergence)] The primary quality metric is the Hamiltonian loss evaluated on binarized solutions, which is the same objective minimized by all three GNN variants. This creates a self-referential evaluation: a method that more aggressively minimizes the QUBO Hamiltonian will look better by construction. The violation counts partly mitigate this for constraints, and the MIS comparison against classical heuristics provides an external check, but for MaxCut and GP there is no ground-truth or independent objective. The paper should acknowledge this limitation explicitly and, if possible, validate on small graphs with known optima or compare solution sizes and partition balances against additional external baselines.
minor comments (5)
  1. [Equation (3)] The map ψω is typed as R → {0,1}, but xGNN is a vector of node assignments; the codomain should be {0,1}^{|V|} (or the map should be described as applied elementwise).
  2. [Algorithm 1, line 7] The line "¯ri_v ← deg(v)P_u∈Vni deg(u) ¯fni" is missing the division operator; it should read deg(v) / Σ_{u∈Vni} deg(u).
  3. [Section 4 (Data Generation)] The number of Louvain levels s and the sizes of each compressed graph are not reported; without these, it is hard to assess the AM load and the claim that all Gi are below the 100k-variable limit.
  4. [Section 5 (Node Decision Assignment Uncertainty)] The caption of Figure 2 contains a typo ("for from a sample"), and the figure's y-axis label is missing.
  5. [Section 5 (Graph Solver Variants)] The text "We omitrGNN" is missing a space, and Table 2's "AM / GNN" notation is not defined in the caption; it should be stated explicitly which side of the slash corresponds to which solver.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the AM-guided initialization is not defined in terms of the final Hamiltonian loss, and the AM labels come from a separate device on compressed graphs.

full rationale

The derivation chain is not circular. AM solutions xAM_i are produced by a CMOS momentum-annealing device on Louvain-compressed graphs Gi (Section 3, Multiresolution Guidance I), not by fitting to the final evaluation metric. The local GNN supervision loss term, ℓ(xGNN_i,t, xAM_i) = ||xGNN_i,t − xAM_i||^2, uses these externally obtained binary vectors, and the final main GNN minimizes HQUBO(xGNN) on the original graph GP. The reported relative loss difference compares two GNN-based solvers that share the same Hamiltonian objective; this shared objective is a methodological property, not a reduction of the claimed result to its inputs. The paper explicitly states its key assumption: 'We assume that while these derived matrices are not equivalent to GP, there should be certain alignment as they are working on different granularities of the original graph GP' (Section 3). This is a stated hypothesis, not a hidden identification. The only notable self-citations are to the authors' own momentum-annealing hardware [23, 30, 31], which support the device's existence and its 100k-variable limit; these are not used to prove the transfer result. The absence of a random-label or heuristic-label control would bear on causal attribution of the gains to the AM specifically, but missing such a control is an experimental-validity concern, not circularity.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim rests on several hand-set hyperparameters (penalty coefficients, learning rates, convergence criteria, AM run settings) and on five domain assumptions: QUBO encoding correctness, Louvain alignment, AM solution quality, loss-as-proxy, and GNN trainability on QUBO objectives. No new physical entities are introduced. The Louvain-alignment assumption is explicitly acknowledged in Section 3; the others are implicit. Because these are empirical inputs rather than derived quantities, the main risk is that the comparison is tuned to the authors' own pipeline rather than to external benchmarks.

free parameters (7)
  • MIS penalty coefficient beta_MIS = 2
    Hand-set in Section 4; controls trade-off between independent set size and edge penalties in HMIS; different values change the objective landscape.
  • Graph Partition penalty coefficient beta_GP = 10
    Hand-set in Section 4 for HGP; balances cut size against partition balance.
  • Node feature dimensionality dK = not stated
    The dimensionality of final node representations after the local GNNs is described as 'predefined' but never reported; it controls the capacity of the transferred representation.
  • Momentum annealing run parameters = 1k variable updates, 1k runs
    Section 4 reports these hand-chosen settings; they affect the quality of AM labels used for supervision.
  • Local loss weighting (Hamiltonian + MSE) = equal weights (implicit)
    Section 4 defines the local loss as the sum of the Hamiltonian and MSE; no weighting or tuning details are given, yet this determines how strongly AM labels shape local GNN features.
  • Learning rates alpha and alpha_in = not stated
    Algorithm 1 takes learning rates as inputs; actual values are not reported, which hampers reproduction.
  • Convergence tolerances = not stated; max epochs 10k main / 1k local
    Section 4 states max epochs but not the tolerance values; these affect when solvers stop and all reported times and losses.
assumptions (5)
  • standard math The QUBO formulations HMaxCut, HMIS, and HGP correctly encode the target optimization problems and their constraints.
    Section 4 states the three Hamiltonians; the approach assumes that minimizing these objectives (with penalty coefficients) yields valid solutions when violations are controlled.
  • domain assumption The Louvain-compressed graphs and their QUBO matrices are sufficiently aligned with the original problem that AM solutions on them provide useful supervision.
    Stated in Section 3 (Multiresolution Guidance I): 'We assume that while these derived matrices are not equivalent to GP, there should be certain alignment as they are working on different granularities of the original graph GP.' This is the load-bearing premise for the whole transfer.
  • domain assumption The annealing machine returns good-quality solutions for the compressed QUBO problems.
    Section 3 uses xAM_i as target labels for local GNNs; no verification of AM solution quality on compressed graphs is reported.
  • domain assumption The Hamiltonian loss value on binarized solutions is a valid proxy for solution quality in the absence of ground truth.
    Section 5 states: 'we used the final loss value as a measure of the quality of the solution, assuming that, in the absence of violations, a lower value means a better solution.' This affects how all comparisons are interpreted.
  • standard math A GNN with soft assignments can be trained by gradient descent on the QUBO Hamiltonian and produce meaningful binary assignments.
    Used throughout the method, inherited from Schuetz et al. [25]; no proof of convergence or optimality is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Annealing Machine-assisted Learning of Graph Neural Network for Combinatorial Optimization." pith.science (2026). https://pith.science/paper/MJW2377Y

@misc{pith2026250105845,
  author       = {Pith},
  title        = {Pith review of: Annealing Machine-assisted Learning of Graph Neural Network for Combinatorial Optimization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MJW2377Y}},
  note         = {Machine review of arXiv:2501.05845}
}
read the original abstract

While Annealing Machines (AM) have shown increasing capabilities in solving complex combinatorial problems, positioning themselves as a more immediate alternative to the expected advances of future fully quantum solutions, there are still scaling limitations. In parallel, Graph Neural Networks (GNN) have been recently adapted to solve combinatorial problems, showing competitive results and potentially high scalability due to their distributed nature. We propose a merging approach that aims at retaining both the accuracy exhibited by AMs and the representational flexibility and scalability of GNNs. Our model considers a compression step, followed by a supervised interaction where partial solutions obtained from the AM are used to guide local GNNs from where node feature representations are obtained and combined to initialize an additional GNN-based solver that handles the original graph's target problem. Intuitively, the AM can solve the combinatorial problem indirectly by infusing its knowledge into the GNN. Experiments on canonical optimization problems show that the idea is feasible, effectively allowing the AM to solve size problems beyond its original limits.

Figures

Figures reproduced from arXiv: 2501.05845 by the authors.

Figure 1
Figure 1. (a) Proposed approach: Original graph GP is compressed into a sequence of decreasing size problem graphs, which are solved by AM. A local GNN solver uses those solutions as guidance, and their resulting learned node representations are aggregated through a Mapping module to initialize a Main GNN solver that attaches to the original graph GP . (b) Guiding block (GB) diagram. Graph Compression We reduce the size of th… view at source ↗
Figure 2
Figure 2. Evolution of the assignment scores from for [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 27 canonical work pages

  1. [1]

    Katzgraber

    Maliheh Aramon, Gili Rosenberg, Elisabetta Valiante, Toshiyuki Miyazawa, Hirotaka Tamura, and Helmut G. Katzgraber. Physics-inspired optimization for quadratic unconstrained problems using a digital Annealer. Front. Phys., 7(APR):48, 2019

  2. [2]

    Fast unfolding of communities in large networks

    Vincent D Blondel, Jean-Loup Guillaume, Renaud Lambiotte, and Etienne Lefebvre. Fast unfolding of communities in large networks. Journal of statistical mechanics: theory and experiment, 2008(10):P10008, 2008

  3. [3]

    Approximating maximum independent sets by excluding subgraphs

    Ravi Boppana and Magnús M Halldórsson. Approximating maximum independent sets by excluding subgraphs. BIT Numerical Mathematics, 32(2):180–196, 1992

  4. [4]

    How attentive are graph attention networks? In International Conference on Learning Representations, 2022

    Shaked Brody, Uri Alon, and Eran Yahav. How attentive are graph attention networks? In International Conference on Learning Representations, 2022

  5. [5]

    Maximum independent set: Self-training through dynamic programming

    Lorenzo Brusca, Lars CPM Quaedvlieg, Stratis Skoulakis, Grigorios Chrysos, and V olkan Cevher. Maximum independent set: Self-training through dynamic programming. Advances in Neural Information Processing Systems, 36, 2024

  6. [6]

    Finding Maximum Cliques on the D-Wave Quantum Annealer

    Guillaume Chapuis, Hristo Djidjev, Georg Hahn, and Guillaume Rizk. Finding Maximum Cliques on the D-Wave Quantum Annealer. J. Signal Process. Syst., 91(3-4):363–377, 2019

  7. [7]

    Clustering approach for solving traveling salesman problems via ising model based solver

    Akira Dan, Riu Shimizu, Takeshi Nishikawa, Song Bian, and Takashi Sato. Clustering approach for solving traveling salesman problems via ising model based solver. In 57th ACM/IEEE Des. Autom. Conf., pages 1–6, 2020

  8. [8]

    Qubo formulations for training machine learning models

    Prasanna Date, Davis Arthur, and Lauren Pusey-Nazzaro. Qubo formulations for training machine learning models. Scientific reports, 11(1):10029, 2021

Show all 33 references
  1. [9]

    Graph communities in neo4j: Four algorithms at work

    Georgios Drakopoulos, Panagiotis Gourgaris, and Andreas Kanavos. Graph communities in neo4j: Four algorithms at work. Evolving Systems, 11(3):397–407, 2020

  2. [10]

    Quantum bridge analytics i: a tutorial on formulating and using qubo models.Annals of Operations Research, 314(1):141–183, 2022

    Fred Glover, Gary Kochenberger, Rick Hennig, and Yu Du. Quantum bridge analytics i: a tutorial on formulating and using qubo models.Annals of Operations Research, 314(1):141–183, 2022. 7

  3. [11]

    High-performance combinatorial optimization based on classical mechanics

    Hayato Goto, Kotaro Endo, Masaru Suzuki, Yoshisato Sakai, Taro Kanao, Yohei Hamakawa, Ryo Hidaka, Masaya Yamasaki, and Kosuke Tatsumura. High-performance combinatorial optimization based on classical mechanics. Sci. Adv., 7(6):eabe7953, 2021

  4. [12]

    Hayato Goto, Kosuke Tatsumura, and Alexander R. Dixon. Combinatorial optimization by simulating adiabatic bifurcations in nonlinear Hamiltonian systems. Sci. Adv., 5(4):eaav2372, 2019

  5. [13]

    Exploring network structure, dynamics, and function using networkx

    Aric Hagberg, Pieter Swart, and Daniel S Chult. Exploring network structure, dynamics, and function using networkx. Technical report, Los Alamos National Lab.(LANL), Los Alamos, NM (United States), 2008

  6. [14]

    Ryan Hamerly, Takahiro Inagaki, Peter L. McMahon, Davide Venturelli, Alireza Marandi, Tatsuhiro Onodera, Edwin Ng, Carsten Langrock, Kensuke Inaba, Toshimori Honjo, Koji Enbutsu, Takeshi Umeki, Ryoichi Kasahara, Shoko Utsunomiya, Satoshi Kako, Ken Ichi Kawarabayashi, Robert L....

  7. [15]

    Graph representation learning

    William L Hamilton. Graph representation learning. Morgan & Claypool Publishers, 2020

  8. [16]

    Enhancing quantum annealing performance for the molecular similarity problem

    Maritza Hernandez and Maliheh Aramon. Enhancing quantum annealing performance for the molecular similarity problem. Quantum Inf. Process., 16(5):133, 2017

  9. [17]

    100,000-spin coherent Ising machine

    Toshimori Honjo, Tomohiro Sonobe, Kensuke Inaba, Takahiro Inagaki, Takuya Ikuta, Ya- suhiro Yamada, Takushi Kazama, Koji Enbutsu, Takeshi Umeki, Ryoichi Kasahara, Ken Ichi Kawarabayashi, and Hiroki Takesue. 100,000-spin coherent Ising machine. Sci. Adv. , 7(40):eabh0952, 2021

  10. [18]

    M. W. Johnson, M. H.S. Amin, S. Gildert, T. Lanting, F. Hamze, N. Dickson, R. Harris, A. J. Berkley, J. Johansson, P. Bunyk, E. M. Chapple, C. Enderud, J. P. Hilton, K. Karimi, E. Ladizinsky, N. Ladizinsky, T. Oh, I. Perminov, C. Rich, M. C. Thom, E. Tolkacheva, C. J.S. Trunci...

  11. [19]

    Accelerating training and inference of graph neural networks with fast sampling and pipelining

    Tim Kaler, Nickolas Stathas, Anne Ouyang, Alexandros-Stavros Iliopoulos, Tao Schardl, Charles E Leiserson, and Jie Chen. Accelerating training and inference of graph neural networks with fast sampling and pipelining. Proceedings of Machine Learning and Systems, 4:172–189, 2022

  12. [20]

    McMahon, and Tim Byrnes

    Naeimeh Mohseni, Peter L. McMahon, and Tim Byrnes. Ising machines as hardware solvers of combinatorial optimization problems. Nat. Rev. Phys., 4(6):363, 2022

  13. [21]

    Dynamic portfolio optimization with real datasets using quantum processors and quantum-inspired tensor networks

    Samuel Mugel, Carlos Kuchkovsky, Escolástico Sánchez, Samuel Fernández-Lorenzo, Jorge Luis-Hita, Enrique Lizaso, and Román Orús. Dynamic portfolio optimization with real datasets using quantum processors and quantum-inspired tensor networks. Phys. Rev. Res., 4(1):013006, 2022

  14. [22]

    Traffic flow optimization using a quantum annealer

    Florian Neukart, Gabriele Compostella, Christian Seidel, David von Dollen, Sheir Yarkoni, and Bob Parney. Traffic flow optimization using a quantum annealer. Front. ICT, 4(DEC):29, 2017

  15. [23]

    Binary optimization by momentum annealing

    Takuya Okuyama, Tomohiro Sonobe, Ken Ichi Kawarabayashi, and Masanao Yamaoka. Binary optimization by momentum annealing. Phys. Rev. E, 100:12111, 2019

  16. [24]

    Solving the Optimal Trading Trajectory Problem Using a Quantum Annealer

    Gili Rosenberg, Poya Haghnegahdar, Phil Goddard, Peter Carr, Kesheng Wu, and Marcos López De Prado. Solving the Optimal Trading Trajectory Problem Using a Quantum Annealer. IEEE J. Sel. Top. Signal Process., 10(6):1053–1060, 2016

  17. [25]

    Combinatorial optimization with physics-inspired graph neural networks

    Martin JA Schuetz, J Kyle Brubaker, and Helmut G Katzgraber. Combinatorial optimization with physics-inspired graph neural networks. Nature Machine Intelligence, 4(4):367–377, 2022

  18. [26]

    A Decomposition Method for Makespan Minimization in Job-Shop Scheduling Problem Using Ising Machine

    Daichi Shimada, Toshiyuki Shibuya, and Takayuki Shibasaki. A Decomposition Method for Makespan Minimization in Job-Shop Scheduling Problem Using Ising Machine. 2021 IEEE 8th Int. Conf. Ind. Eng. Appl. ICIEA 2021, page 307, 2021

  19. [27]

    Graph neural networks for maximum constraint satisfaction

    Jan Toenshoff, Martin Ritzert, Hinrikus Wolf, and Martin Grohe. Graph neural networks for maximum constraint satisfaction. Frontiers in artificial intelligence, 3:580607, 2021

  20. [28]

    An accelerator architecture for combinatorial optimization problems

    Sanroku Tsukamoto, Motomu Takatsu, Satoshi Matsubara, and Hirotaka Tamura. An accelerator architecture for combinatorial optimization problems. Fujitsu Sci. Tech. J, 53(5):8–13, 2017. 8

  21. [29]

    Graph attention networks

    Petar Veliˇckovi´c, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Liò, and Yoshua Bengio. Graph attention networks. In International Conference on Learning Representations, 2018

  22. [30]

    A 1.3-Mbit Annealing System Composed of Fully-Synchronized 9-board x 9-chip x 16-kbit Annealing Processor Chips for Large-Scale Combinatorial Optimization Problems

    Kasho Yamamoto, Takashi Takemoto, Chihiro Yoshimura, Mayumi Mashimo, and Masanao Yamaoka. A 1.3-Mbit Annealing System Composed of Fully-Synchronized 9-board x 9-chip x 16-kbit Annealing Processor Chips for Large-Scale Combinatorial Optimization Problems. Proc. - A-SSCC 2021 IE...

  23. [31]

    20k-spin Ising chip for combinational optimization problem with CMOS annealing

    Masanao Yamaoka, Chihiro Yoshimura, Masato Hayashi, Takuya Okuyama, Hidetaka Aoki, and Hiroyuki Mizuno. 20k-spin Ising chip for combinational optimization problem with CMOS annealing. Dig. Tech. Pap. - IEEE Int. Solid-State Circuits Conf., 58:432, 2015

  24. [32]

    QUBO-inspired Molecular Fingerprint for Chemical Property Prediction

    Koichiro Yawata, Yoshihiro Osakabe, Takuya Okuyama, and Akinori Asahara. QUBO-inspired Molecular Fingerprint for Chemical Property Prediction. Proc. - 2022 IEEE Int. Conf. Big Data, Big Data 2022, pages 2437–2440, 2022

  25. [33]

    Distdgl: distributed graph neural network training for billion-scale graphs

    Da Zheng, Chao Ma, Minjie Wang, Jinjing Zhou, Qidong Su, Xiang Song, Quan Gan, Zheng Zhang, and George Karypis. Distdgl: distributed graph neural network training for billion-scale graphs. In 2020 IEEE/ACM 10th Workshop on Irregular Applications: Architectures and Algorithms (...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.