Pith. sign in

REVIEW 3 major objections 5 minor 27 references

Rank-Refined Quantum-Behaved Particle Swarm Optimization for Quantum Molecular Generation

T0 review · 3 major / 5 minor · reviewed 2026-07-14 · grok-4.5

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

Pith's one-line read A rank-refined quantum particle-swarm optimizer raises quantum molecular generation quality without changing the chemistry circuit.

desk verdict Clean empirical optimizer paper for QMG: modest, consistent gains over re-run BO, but single-seed and hand-tuned, so treat the 4-point claim as provisional. read the letter →

arxiv 2607.10284 v1 pith:B5VTOU3P submitted 2026-07-11 quant-ph

Sing-Yun Wu , Sheng Yun Wu , I-Min Chiang , Tai-Yue Li This is my paper
classification quant-ph
keywords QuantumMolecularGenerationQuantum-BehavedParticleSwarmOptimizationMulti-objectiveHybridQuantum-ClassicalComputingparametervalidity-uniquenessproductdynamiccircuit
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

Quantum molecular generation draws candidate molecules from a chemistry-inspired dynamic quantum circuit whose parameters must be tuned by costly black-box evaluations: each trial needs many circuit shots, bitstring decoding, and chemical checks. This paper claims that a population method, Rank-Refined Quantum-Behaved Particle Swarm Optimization, can replace Bayesian optimization for that search. It combines low-discrepancy Sobol start points, a rank-corrected mean-best attractor that separates strong and weak personal bests, and fitness-guided elites that track validity and uniqueness separately. On the fixed 9-heavy-atom, 134-parameter, 20-qubit benchmark the method lifts the validity–uniqueness product from 0.902 for Bayesian optimization to 0.930 with 64 particles and 0.942 with 128 particles under the same scoring protocol, and it keeps a higher quality product when the objective is also steered toward target hydrogen-bond counts. A reader who wants better molecules without redesigning the quantum circuit or the decoder therefore has a concrete classical lever: redesign the optimizer that sits around the expensive sampling pipeline.

What carries the argument

Rank-Refined Quantum-Behaved Particle Swarm Optimization (RR-QPSO): candidate circuit-parameter vectors are moved by a quantum-behaved position update whose swarm mean-best is first shifted by a rank correction that pulls high-fitness personal bests away from low-fitness ones, then mixed with validity- and uniqueness-oriented elite attractors; this population signal replaces a Bayesian surrogate for the expensive, noisy objective.

What would settle it

A multi-seed comparison that uses the same total evaluation budget and shows Bayesian optimization matching or exceeding the final validity–uniqueness distribution of RR-QPSO on the 9-heavy-atom benchmark would falsify the claimed optimizer advantage.

Watch

Extended reading notes

Core claim

On the 9-heavy-atom Quantum Molecular Generation benchmark, Rank-Refined Quantum-Behaved Particle Swarm Optimization reaches a validity–uniqueness product of 0.942 with 128 particles and 150 iterations, against 0.902 for Bayesian optimization under the identical circuit, decoding, and 5000-shot scoring protocol. The improvement is attributed solely to classical optimizer design—Sobol initialization, a rank-refined mean-best update, and validity- and uniqueness-aware elite attractors—without any change to the chemistry-inspired 20-qubit circuit or molecular decoder. Under a scalarized multi-objective objective that also targets hydrogen-bond acceptor and donor counts near 4 and 3, the same sw

Load-bearing premise

The superiority claim rests on single-seed, fixed-budget comparisons that the paper itself calls benchmark-level evidence rather than a complete statistical ranking.

Editorial extensions

If this is right

  • Optimizer-level redesign alone can raise the validity–uniqueness product of an unchanged QMG circuit and decoder.
  • Larger independent particle counts, evaluated in parallel, systematically improve final scores under a fixed iteration budget.
  • The same swarm can absorb scalarized property targets (HBA/HBD) while preserving higher molecular quality than Bayesian search in the reported setting.
  • When each candidate evaluation is expensive yet independent, population search becomes a natural fit for multi-GPU QMG workflows.

Reading between the lines

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

  • If the rank-refined attractor generalizes, similar swarm corrections may help other high-dimensional variational quantum tuning problems whose objectives are noisy sample-based metrics.
  • Embarrassingly parallel particle evaluations suggest wall-clock advantage for swarm methods over sequential Bayesian proposals as shot counts or molecular size grow.
  • Property histograms or Pareto fronts, noted as future work by the authors, would show whether mean-centered scalarization truly concentrates molecules near the target region or only averages to it.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The manuscript proposes Rank-Refined Quantum-Behaved Particle Swarm Optimization (RR-QPSO) as a population-based alternative to Bayesian optimization for tuning the 134 parameters of a chemistry-inspired dynamic quantum circuit used in Quantum Molecular Generation (QMG). The method combines Owen-scrambled Sobol initialization, a rank-refined mean-best position (Eq. 9) that biases the swarm attractor toward high-fitness personal bests, and a fitness-guided attractor (Eq. 10) that mixes validity- and uniqueness-oriented elites under complementary-metric thresholds. On the 9-heavy-atom, 20-qubit CUDA-Q benchmark with 5000-shot evaluations, RR-QPSO reports V×U = 0.930 (M=64) and 0.942 (M=128) versus 0.902 for a re-run BO baseline under the same circuit, decoding, and scoring pipeline; a scalarized multi-objective extension targeting HBA=4 and HBD=3 likewise retains a higher validity–uniqueness product while centering mean property counts near the targets. The authors argue that optimizer-level design can improve QMG without changing the circuit or molecular decoder, and that independent particle evaluations make the approach naturally multi-GPU parallelizable.

Significance. If the reported gains hold under stronger statistical controls, the work is a useful systems-level contribution at the intersection of quantum generative modeling and expensive black-box optimization: it shows that a carefully adapted QPSO variant can outperform a standard BO baseline on a published QMG circuit without redesigning the chemistry-inspired ansatz or the decoding pipeline. Strengths include a clean external comparison protocol (identical circuit, 5000-shot scoring, re-implemented BO), explicit multi-GPU parallel evaluation of independent particles, and a demonstrated transfer of the same optimizer to a scalarized property-guided objective. The paper is also candid that the present numbers are “benchmark-level evidence rather than a complete statistical ranking.” The main scientific value is therefore practical—optimizer design for stochastic, high-dimensional QMG parameter search—rather than a new quantum generative model or a theoretical guarantee.

major comments (3)
  1. [Section IV-A/B, Fig. 2–3, Table I] Section IV-A/B, Fig. 2–3, and Table I: the central superiority claim (V×U = 0.942 for RR-QPSO at M=128 vs 0.902 for BO) rests on single-seed trajectories under a noisy 5000-shot objective. Because F(θ) is stochastic, a single swarm path can overstate the method gap. The manuscript itself labels the comparison “benchmark-level evidence rather than a complete statistical ranking.” Multi-seed means, standard deviations (or confidence intervals), and ideally a fixed total-evaluation-budget comparison against BO are needed before the 4-point absolute gain can be treated as a reliable optimizer effect rather than a plausible but unquantified trend.
  2. [Section III-C/D, Fig. 2] Section III-C/D and the free-parameter set (ρ=0.015, wRR=0.70, wV=wU=0.15, τV=τU=0.5, αmax=1.2, αmin=0.3): these knobs are fixed on the same 9-heavy-atom benchmark used for the headline comparison, yet no sensitivity or leave-one-component-out ablation is reported beyond the coarse BO / QPSO / QPSO+Sobol / RR-QPSO ladder in Fig. 2. At minimum, a short sensitivity study on ρ and the elite weights (or an ablation that isolates rank refinement from fitness-guided elites) is required to show that the gain is not an artifact of hand-tuning on the reported seed.
  3. [Section IV-A] Section IV-A and the BO baseline protocol: the text states that BO is “re-run under the same implementation and evaluation protocol,” but does not specify the BO acquisition function, surrogate, number of initial points, or total number of objective evaluations relative to M×T particle evaluations (e.g., 64×150 = 9600 or 128×150 = 19200). Without an equal- or matched-budget statement, the comparison risks confounding optimizer quality with search effort. Please state the BO evaluation budget and, if it differs, either re-run BO at a matched budget or report both wall-clock and evaluation-count comparisons explicitly.
minor comments (5)
  1. [Section II, Eq. (1)] Equation (1) for D is written as D = 8 + 3(N−2)(N+3)/2; a brief derivation or citation pointer to the original QMG circuit paper would help readers verify D=134 for N=9 without external lookup.
  2. [Fig. 2, Abstract, Table I] Figure 2 reports percentages (e.g., 93.0) while the abstract and Table I mix 0.930 / 0.942 and percent forms; unify V×U reporting to one convention throughout.
  3. [Table I] Table I: runtime is non-monotonic in M (e.g., M=96 at 43.67 h vs M=64 at 47.12 h). A one-sentence note on worker scheduling or load imbalance would prevent misreading the table as a scaling anomaly.
  4. [Section IV-C, Fig. 4] Section IV-C multi-objective experiment: only M=16 and M=32 are shown against BO, whereas the unconditional study goes to M=128. A brief justification for the smaller swarms (or one larger-M curve) would strengthen the multi-objective claim.
  5. [Fig. 1, Fig. 2] Typos / wording: “Iterate until convergence” in Fig. 1 is slightly at odds with the fixed T=150 budget used in experiments; “QPSO + Sobol Init.” in Fig. 2 could be labeled consistently with the text (“QPSO with Sobol initialization”).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: empirical optimizer comparison against a re-run BO baseline on an independently defined V imes U metric and published QMG circuit.

full rationale

The paper is a methods/empirical optimization paper, not a first-principles derivation. Fitness F(θ)=V(θ)×U(θ) is defined from finite circuit samples, decoding, and chemical validity checks (Eqs. 2–3); RR-QPSO is a population update (Sobol init, rank-refined mean-best Eq. 9 with fixed ρ=0.015, fitness-guided attractor Eq. 10) that is then run against a re-implemented BO baseline under the identical 134-parameter/20-qubit circuit, 5000-shot protocol, and scoring pipeline. The reported gains (0.930/0.942 vs 0.902) are experimental outcomes of that comparison, not quantities forced by construction from fitted constants or self-referential definitions. Multi-objective scalarization (Eq. 12) is an explicit weighted objective, not a hidden tautology. Self-citations ([22] for the QMG circuit, classic QPSO refs) supply background and the evaluation pipeline; they do not load-bear the superiority claim, which rests on the authors’ own parallel GPU runs. Single-seed/fixed-budget limitations affect statistical strength but are not circularity. No equation reduces a claimed prediction to its inputs; the derivation chain is self-contained against the external benchmark.

Assumptions & free parameters 4 free parameters · 4 assumptions · 2 invented entities

The central empirical claim rests on a small set of hand-chosen swarm hyper-parameters, the standard QPSO probabilistic update, the published QMG circuit definition, and the assumption that single-seed V×U comparisons under fixed evaluation budget are informative. No new physical entities are postulated; free parameters are optimizer knobs rather than fitted physical constants.

free parameters (4)
  • rank-correction strength ρ = 0.015
    Fixed at 0.015 for all iterations; controls how strongly high-fitness personal bests pull the mean-best attractor (Eq. 9). Chosen by hand, not derived.
  • elite weights and thresholds (wRR, wV, wU, τV, τU) = wRR=0.70, wV=wU=0.15, τ=0.5
    Set to 0.70 / 0.15 / 0.15 and 0.5 / 0.5; determine when validity- or uniqueness-oriented elites enter the guided attractor (Eqs. 10–11). Hand-tuned.
  • contraction–expansion bounds αmax, αmin = 1.2 / 0.3
    Bounded to [0.3, 1.2] to control exploration versus exploitation schedule; standard QPSO practice but still free.
  • multi-objective scalarization weight λ and property sigmas = λ=0.40, σ=1
    λ=0.40, σHBA=σHBD=1 control the trade-off between V×U and property closeness (Eqs. 12–13); chosen for the reported experiment.
assumptions (4)
  • domain assumption Standard QPSO position update (local attractor + logarithmic random step around mean-best) is a valid black-box optimizer for noisy, high-dimensional continuous objectives.
    Invoked throughout Section III-C; taken from Sun et al. without re-derivation.
  • domain assumption The 9-heavy-atom QMG circuit (D=134, 20 qubits) and its validity/uniqueness scoring pipeline correctly measure molecular generation quality.
    Adopted wholesale from Chen et al. (2025); all comparisons rest on this fixed black-box (Section II).
  • domain assumption Candidate evaluations are independent and therefore perfectly parallelizable across GPUs without communication or bias.
    Stated in Sections I and III-E; underpins the multi-GPU scaling argument.
  • standard math Owen-scrambled Sobol sequences provide superior space-filling initial coverage relative to pure random initialization in 134 dimensions.
    Cited from Sobol/Owen; used for swarm initialization (Section III-B).
invented entities (2)
  • Rank-refined mean-best position m_RR
    purpose: Replaces the ordinary average of personal bests with a rank-weighted correction that emphasizes the difference between high- and low-fitness elites (Eq. 9).
    Defined ad hoc for this paper; no independent theoretical guarantee or external validation is supplied beyond the reported QMG runs.
  • Fitness-guided attractor combining validity- and uniqueness-oriented elites
    purpose: Provides auxiliary search directions that balance the product objective’s two failure modes (Eqs. 10–11).
    Constructed specifically for the V×U objective of QMG; evidence is internal to the experiments.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Rank-Refined Quantum-Behaved Particle Swarm Optimization for Quantum Molecular Generation." pith.science (2026). https://pith.science/paper/B5VTOU3P

@misc{pith2026260710284,
  author       = {Pith},
  title        = {Pith review of: Rank-Refined Quantum-Behaved Particle Swarm Optimization for Quantum Molecular Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B5VTOU3P}},
  note         = {Machine review of arXiv:2607.10284}
}
read the original abstract

This work proposes Rank-Refined Quantum-Behaved Particle Swarm Optimization (RR-QPSO) for high-dimensional parameter search in Quantum Molecular Generation (QMG). RR-QPSO targets the optimization bottleneck caused by expensive objective evaluations, where each candidate parameter vector requires stochastic circuit sampling, bitstring decoding, and molecular evaluation. The method provides a population-based alternative to Bayesian optimization (BO), combining Sobol-based initialization, a rank-refined mean-best update, and fitness-guided refinement based on validity and uniqueness. Experiments use the 9-heavy-atom QMG benchmark with a 134-parameter, 20-qubit CUDA-Q circuit and particle evaluations parallelized across 8 NVIDIA V100 GPUs. With M=64 particles and T=150 iterations, RR-QPSO reaches VxU = 0.930; increasing the swarm size to M=128 further improves the product to 0.942, compared with 0.902 for BO under the same protocol. A multi-objective extension targeting HBA=4 and HBD=3 further shows that RR-QPSO can guide molecular properties while preserving a higher validity--uniqueness product than BO. These results suggest that optimizer-level design can improve QMG without modifying the chemistry-inspired circuit or molecular decoding pipeline.

Figures

Figures reproduced from arXiv: 2607.10284 by the authors.

Figure 1
Figure 1. RR-QPSO workflow for QMG parameter optimization. Sobol sampling initializes a swarm of [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Optimizer performance comparison on the 9-heavy-atom QMG [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Multi-objective optimization targeting HBA = 4 and HBD = 3. BO and RR-QPSO are evaluated under the same scalarized target-property objective. RR-QPSO retains a higher validity–uniqueness product while keeping the generated molecules close to the target HBA and HBD region in this scalarized setting. bond donor (HBD) counts, with desired values HBA = 4 and HBD = 3. This setting introduces a trade-off between molecular… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

27 extracted references · 5 linked inside Pith

  1. [1]

    Estimation of the size of drug-like chemical space based on GDB-17 data,

    P. G. Polishchuk, T. I. Madzhidov, and A. Varnek, “Estimation of the size of drug-like chemical space based on GDB-17 data,”Journal of Computer-Aided Molecular Design, vol. 27, no. 8, pp. 675–679, 2013

    2013

  2. [2]

    Automatic chemical design using a data-driven continuous representation of molecules,

    R. G ´omez-Bombarelli, J. N. Wei, D. Duvenaud, J. M. Hern ´andez- Lobato, B. S ´anchez-Lengeling, D. Sheberla, J. Aguilera-Iparraguirre, T. D. Hirzel, R. P. Adams, and A. Aspuru-Guzik, “Automatic chemical design using a data-driven continuous representation of molecules,”ACS Central Science, vol. 4, no. 2, pp. 268–276, 2018

    2018

  3. [3]

    Junction tree variational autoen- coder for molecular graph generation,

    W. Jin, R. Barzilay, and T. Jaakkola, “Junction tree variational autoen- coder for molecular graph generation,” inInternational Conference on Machine Learning (ICML), 2018

    2018

  4. [4]

    Generating focused molecule libraries for drug discovery with recurrent neural networks,

    M. H. S. Segler, T. Kogej, C. Tyrchan, and M. P. Waller, “Generating focused molecule libraries for drug discovery with recurrent neural networks,”ACS Central Science, vol. 4, no. 1, pp. 120–131, 2018

    2018

  5. [5]

    MolGAN: An implicit generative model for small molecular graphs,

    N. De Cao and T. Kipf, “MolGAN: An implicit generative model for small molecular graphs,”arXiv preprint arXiv:1805.11973, 2018

    arXiv 2018

  6. [6]

    Quantum machine learning,

    J. Biamonte, P. Wittek, N. Pancotti, P. Rebentrost, N. Wiebe, and S. Lloyd, “Quantum machine learning,”Nature, vol. 549, no. 7671, pp. 195–202, 2017

    2017

  7. [7]

    Variational quantum algorithms,

    M. Cerezo, A. Arrasmith, R. Babbush, S. C. Benjamin, S. Endo, K. Fujii, J. R. McClean, K. Mitarai, X. Yuan, L. Cincio,et al., “Variational quantum algorithms,”Nature Reviews Physics, vol. 3, no. 9, pp. 625– 644, 2021

    2021

  8. [8]

    Quantum support vector machine for big data classification,

    P. Rebentrost, M. Mohseni, and S. Lloyd, “Quantum support vector machine for big data classification,”Physical review letters, vol. 113, no. 13, p. 130503, 2014

    2014

Show all 27 references
  1. [9]

    Supervised learning with quantum- enhanced feature spaces,

    V . Havl ´ıˇcek, A. D. C ´orcoles, K. Temme, A. W. Harrow, A. Kandala, J. M. Chow, and J. M. Gambetta, “Supervised learning with quantum- enhanced feature spaces,”Nature, vol. 567, no. 7747, pp. 209–212, 2019

    2019

  2. [10]

    Classification of tumor metastasis data by using quantum kernel-based algorithms,

    T.-Y . Li, V . R. Mekala, K.-L. Ng, and C.-F. Su, “Classification of tumor metastasis data by using quantum kernel-based algorithms,” in2022 IEEE 22nd International Conference on Bioinformatics and Bioengineering (BIBE), pp. 351–354, IEEE, 2022

    2022

  3. [11]

    Quantum pointwise convolution: A flexible and scalable approach for neural network enhancement,

    A. Ning, T.-Y . Li, and N.-Y . Chen, “Quantum pointwise convolution: A flexible and scalable approach for neural network enhancement,” in2025 International Conference on Quantum Communications, Networking, and Computing (QCNC), pp. 371–378, IEEE, 2025

    2025

  4. [12]

    Quantum adaptive excitation network with variational quantum circuits for channel atten- tion,

    Y .-C. Hsu, K.-C. Chen, T.-Y . Li, and N.-Y . Chen, “Quantum adaptive excitation network with variational quantum circuits for channel atten- tion,”arXiv preprint arXiv:2507.11217, 2025

    2025 arXiv

  5. [13]

    Hybrid classical-quantum neural networks for multi-characteristic co- optimization of recessed-gate AlGaN/GaN MIS-HEMTs,

    R. Rai, P.-J. Chang, D. V . Nguyen, Y .-C. Chiu, N. Tumilty, Y .- Y . Wang, S. See, W.-J. Lee, T.-Y . Li, N.-Y . Chen, and T.-L. Wu, “Hybrid classical-quantum neural networks for multi-characteristic co- optimization of recessed-gate AlGaN/GaN MIS-HEMTs,”arXiv preprint arXiv:2...

    2026 arXiv

  6. [14]

    MPM- QIR: Measurement-probability matching for quantum image represen- tation and compression via variational quantum circuit,

    C.-W. Wang, M. I. Sam, T.-L. Kuo, N.-Y . Chen, and T.-Y . Li, “MPM- QIR: Measurement-probability matching for quantum image represen- tation and compression via variational quantum circuit,”arXiv preprint arXiv:2601.03855, 2026

    2026

  7. [15]

    Validating large-scale quantum ma- chine learning: Efficient simulation of quantum support vector machines using tensor networks,

    K.-C. Chen, T.-Y . Li, Y .-Y . Wang, S. See, C.-C. Wang, R. Wille, N.-Y . Chen, A.-C. Yang, and C.-Y . Lin, “Validating large-scale quantum ma- chine learning: Efficient simulation of quantum support vector machines using tensor networks,”Machine Learning: Science and Technology, 2024

    2024

  8. [16]

    Scalable tensor network simulation for quantum-classical dual kernel,

    M. I. Sam and T.-Y . Li, “Scalable tensor network simulation for quantum-classical dual kernel,”arXiv preprint arXiv:2602.01330, 2026

    2026

  9. [17]

    Iterative matrix product state simu- lation for scalable grover’s algorithm,

    M. I. Sam, T.-L. Kuo, and T.-Y . Li, “Iterative matrix product state simu- lation for scalable grover’s algorithm,”arXiv preprint arXiv:2601.03832, 2026

    2026

  10. [18]

    Quantum generative adversarial networks,

    P.-L. Dallaire-Demers and N. Killoran, “Quantum generative adversarial networks,”Physical Review A, vol. 98, no. 1, p. 012324, 2018

    2018

  11. [19]

    Differentiable learning of quantum circuit Born machines,

    J.-G. Liu and L. Wang, “Differentiable learning of quantum circuit Born machines,”Physical Review A, vol. 98, no. 6, p. 062324, 2018

    2018

  12. [20]

    Drug discovery approaches using quantum machine learning,

    J. Li, M. Alam, C. M. Sha, J. Wang, N. V . Dokholyan, and S. Ghosh, “Drug discovery approaches using quantum machine learning,”arXiv preprint arXiv:2104.00746, 2021

    2021 arXiv

  13. [21]

    QCA-MolGAN: Quantum circuit associative molecular GAN with multi-agent reinforcement learning,

    A. M. Thomas, Y .-C. Chen, H. O. Valencia, S. T. Jose, and R. Wu, “QCA-MolGAN: Quantum circuit associative molecular GAN with multi-agent reinforcement learning,”arXiv preprint arXiv:2509.05051, 2025

    2025 arXiv

  14. [22]

    Exploring chemical space with chemistry-inspired dynamic quantum circuits in the nisq era,

    L.-Y . Chen, T.-Y . Li, Y .-P. Li, N.-Y . Chen, and F. You, “Exploring chemical space with chemistry-inspired dynamic quantum circuits in the nisq era,”Journal of Chemical Theory and Computation, 2025

    2025

  15. [23]

    Practical bayesian optimiza- tion of machine learning algorithms,

    J. Snoek, H. Larochelle, and R. P. Adams, “Practical bayesian optimiza- tion of machine learning algorithms,” inAdvances in Neural Information Processing Systems, vol. 25, pp. 2951–2959, 2012

    2012

  16. [24]

    On the distribution of points in a cube and the approx- imate evaluation of integrals,

    I. M. Sobol’, “On the distribution of points in a cube and the approx- imate evaluation of integrals,”USSR Computational Mathematics and Mathematical Physics, vol. 7, no. 4, pp. 86–112, 1967

    1967

  17. [25]

    Randomly permuted(t, m, s)-nets and(t, s)-sequences,

    A. B. Owen, “Randomly permuted(t, m, s)-nets and(t, s)-sequences,” inMonte Carlo and Quasi-Monte Carlo Methods in Scientific Com- puting, vol. 106 ofLecture Notes in Statistics, pp. 299–317, Springer, 1995

    1995

  18. [26]

    Particle swarm optimization with particles having quantum behavior,

    J. Sun, B. Feng, and W. Xu, “Particle swarm optimization with particles having quantum behavior,” inProceedings of the 2004 Congress on Evolutionary Computation, vol. 1, pp. 325–331, IEEE, 2004

    2004

  19. [27]

    Quantum-behaved particle swarm optimization: analysis of individual particle behavior and parameter selection,

    J. Sun, W. Fang, X. Wu, V . Palade, and W. Xu, “Quantum-behaved particle swarm optimization: analysis of individual particle behavior and parameter selection,”Evolutionary Computation, vol. 20, no. 3, pp. 349– 393, 2012

    2012

Pith tools

Reviewed July 14, 2026 · model on record in the stance chip above.