pith. machine review for the scientific record. sign in

arxiv: 2605.13268 · v1 · submitted 2026-05-13 · 🪐 quant-ph · cs.LG

Recognition: no theorem link

Physics Guided Generative Optimization for Trotter Suzuki Decomposition

WenBin Yan
Authors on Pith no claims yet

Pith reviewed 2026-05-14 18:36 UTC · model grok-4.3

classification 🪐 quant-ph cs.LG
keywords Trotter-Suzuki decompositionNISQ compilationgenerative optimizationphysics-informed neural networksquantum simulationdiffusion modelstransverse field Ising modelcircuit depth reduction
0
0 comments X

The pith

A conditional diffusion model guided by physics-informed fidelity feedback produces Trotter-Suzuki decompositions that reach 85.6 percent of fourth-order baseline accuracy at 22 percent circuit depth on the transverse Ising model.

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

The paper presents a closed generate-and-evaluate loop that automates the coupled choices of term grouping, product-formula order, and timestep allocation in Trotter-Suzuki formulas. A conditional diffusion model proposes candidate strategies in a mixed discrete-continuous space, a physics-informed neural network supplies differentiable estimates of simulation fidelity, and a graph neural network encodes the commutator structure among Hamiltonian terms. These components are trained together with REINFORCE gradients and a Pareto tracker that balances depth against accuracy. On the transverse field Ising model the resulting decompositions recover 85.6 percent of the fidelity of a standard fourth-order Qiskit baseline while using only 21.8 percent of the circuit depth and 19.2 percent of the CNOT count. When depth is held fixed to the baseline budget, further fine-tuning inside the loop reaches observed fidelities of 0.9994.

Core claim

The central claim is that a hybrid optimization loop combining a conditional diffusion model for proposing decomposition strategies, a physics-informed neural network for differentiable fidelity supervision, and a graph neural network for commutator encoding can generate Trotter-Suzuki product formulas whose accuracy-to-depth trade-off on the transverse field Ising model substantially exceeds that of hand-tuned baselines such as fourth-order Qiskit.

What carries the argument

The generate-and-evaluate loop that uses a conditional diffusion model to sample hybrid discrete-continuous decomposition strategies, guided by fidelity gradients from a physics-informed neural network and commutator structure from a graph neural network, trained via REINFORCE with Pareto tracking.

If this is right

  • Shallow circuits become practical for Hamiltonian simulation on NISQ hardware without sacrificing most of the target accuracy.
  • Manual heuristics for term grouping and order selection can be replaced by automated search under a fixed training budget.
  • Explicit Pareto control allows users to trade depth for fidelity according to hardware limits.
  • Fine-tuning inside the loop can push performance near unity when depth is not the binding constraint.
  • The same supervision signal may transfer to other Hamiltonians whose commutator graphs are similar.

Where Pith is reading between the lines

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

  • If the PINN fidelity predictor remains accurate for larger or non-Ising Hamiltonians, the method could automate compilation for models where exhaustive search is impossible.
  • The loop could be extended to incorporate device-specific noise models directly into the physics-informed feedback.
  • Combining the diffusion proposer with other generative architectures might lower the data or compute needed for training.
  • The same generate-evaluate pattern might apply to other discrete-continuous compilation problems such as gate scheduling or ansatz design.

Load-bearing premise

The physics-informed neural network must continue to give reliable differentiable fidelity estimates across the discrete groupings and formula orders that the diffusion model explores during training.

What would settle it

Run the reported optimized decompositions on the transverse field Ising model and check whether they achieve at least 0.85 fidelity at the stated 21.8 percent depth and 19.2 percent CNOT reduction relative to the fourth-order Qiskit baseline.

Figures

Figures reproduced from arXiv: 2605.13268 by WenBin Yan.

Figure 1
Figure 1. Figure 1: Marginal statistics for the TFIM training corpus. Left to right: qubit count [PITH_FULL_IMAGE:figures/full_fig_p010_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: PINN validation on ten TFIM instances. Left: post-training PDE residual vs. 10 [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Bar-style comparison (fidelity, depth, latency on a log axis). Fidelity trails the high [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Fidelity across 50 closed-loop iterations. No collapses; variance reflects REINFORCE [PITH_FULL_IMAGE:figures/full_fig_p012_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Example grouping heatmap. Imbalanced groups emerge naturally: singletons with [PITH_FULL_IMAGE:figures/full_fig_p013_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Mean fidelity vs. n for Heisenberg instances (blue: ours, green: Qiskit fourth). Depth stays ∼8× smaller, but 50-step adaptation underfits relative to baselines as n grows. 14 [PITH_FULL_IMAGE:figures/full_fig_p014_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Molecular scans. Left: H2 exact fidelity (26 bonds); our curve (blue) tracks the Qiskit fourth-order baseline (green) with ∼11.2% depth. Right: LiH proxy fidelity (31 bonds); ours stays above the baseline despite aggressive depth cuts. LiH: twelve qubits and 631 Pauli strings push Qiskit fourth order to depth 6305—impractical on many machines. At depth 812.8 our proxy fidelity is 0.9603 vs. 0.7296 for the … view at source ↗
read the original abstract

Product formulas for Trotter Suzuki simulation remain a practical route to Hamiltonian evolution on noisy intermediate scale quantum (NISQ) hardware, yet their accuracy hinges on three coupled choices: term grouping, product formula order, and timestep allocation. Toolchains such as Qiskit and Paulihedral lean on hand tuned heuristics, while the discrete nature of grouping and order makes naive gradient based optimization awkward. We describe a generate and evaluate loop: a conditional diffusion model proposes strategies, a physics informed neural network (PINN) supplies differentiable fidelity feedback, and a graph neural network (GNN) encodes commutator structure. Training spans a hybrid space (discrete grouping and order, continuous time steps); the closed loop uses REINFORCE and a Pareto tracker. On the transverse field Ising model (TFIM), under our primary comparison setup, the method reaches 85.6% of the fidelity of a fourth order Qiskit baseline (0.856) at roughly 21.8% of the circuit depth and 19.2% of the baseline CNOT count. Under an equal depth budget, fine tuning in the loop reached a best observed fidelity of 0.9994. Updated ablations show that, for a fixed training budget and default guidance knobs, module contributions depend on the training recipe and guidance hyperparameters CFG in particular needs to be tuned jointly with compute budget. Overall, the results suggest that "generative model and physics supervision" is a viable angle for NISQ oriented compilation, though where it wins still depends on the operating point.

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

2 major / 2 minor

Summary. The paper proposes a generate-and-evaluate framework for optimizing Trotter-Suzuki decompositions that combines a conditional diffusion model to propose discrete groupings and orders together with continuous timesteps, a physics-informed neural network (PINN) to supply differentiable fidelity feedback, and a graph neural network (GNN) to encode commutator structure. Training proceeds via a closed REINFORCE loop with a Pareto tracker. On the transverse-field Ising model the method is reported to reach 85.6 % of the fidelity of a fourth-order Qiskit baseline at 21.8 % circuit depth and 19.2 % CNOT count, with equal-depth fine-tuning attaining a best observed fidelity of 0.9994.

Significance. If the PINN surrogate remains accurate for the discrete structural choices explored by the diffusion model and the learned strategies generalize, the approach would constitute a viable data-driven alternative to hand-tuned heuristics for NISQ-oriented product-formula compilation. The hybrid discrete-continuous generative optimization and the explicit use of physics supervision are technically interesting and could be extended to other Hamiltonian simulation tasks.

major comments (2)
  1. [Results on TFIM and ablation studies] The headline performance figures (85.6 % relative fidelity at reduced depth, 0.9994 under equal-depth fine-tuning) are obtained from a closed REINFORCE loop whose reward is the PINN fidelity estimate. No section reports a held-out validation of this estimate against exact Trotter fidelity (or high-fidelity reference simulation) across the discrete grouping and order axes that the diffusion model explores. Without this check the optimizer may exploit surrogate artifacts rather than genuine improvements.
  2. [Abstract and experimental results] The abstract and results claim concrete fidelity and resource numbers (0.856 relative fidelity, 21.8 % depth, 19.2 % CNOT) yet provide neither error bars nor a description of the training-data distribution over Hamiltonian instances and hyper-parameters. This omission makes it impossible to assess the statistical reliability of the reported gains.
minor comments (2)
  1. [Ablation studies] The updated ablations note that module contributions depend on training recipe and CFG guidance scale, but the manuscript does not quantify how these dependencies affect generalization beyond the specific TFIM instances used.
  2. [Methods] Notation for the conditional diffusion model, the PINN architecture, and the GNN commutator encoder should be introduced with explicit equations or diagrams to improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major point below and indicate the revisions that will be incorporated in the updated version.

read point-by-point responses

  1. Referee: [Results on TFIM and ablation studies] The headline performance figures (85.6 % relative fidelity at reduced depth, 0.9994 under equal-depth fine-tuning) are obtained from a closed REINFORCE loop whose reward is the PINN fidelity estimate. No section reports a held-out validation of this estimate against exact Trotter fidelity (or high-fidelity reference simulation) across the discrete grouping and order axes that the diffusion model explores. Without this check the optimizer may exploit surrogate artifacts rather than genuine improvements.

    Authors: We agree that the absence of an explicit held-out validation of the PINN surrogate against exact fidelity on the discrete structural choices is a limitation that should be addressed. The current manuscript uses the PINN reward inside the closed training loop but does not report an independent comparison on held-out configurations. In the revised manuscript we will add a dedicated subsection that computes exact Trotter fidelities (via direct high-fidelity simulation) for 500 held-out discrete groupings and orders sampled from the diffusion model and reports the correlation with the PINN estimates. revision: yes

  2. Referee: [Abstract and experimental results] The abstract and results claim concrete fidelity and resource numbers (0.856 relative fidelity, 21.8 % depth, 19.2 % CNOT) yet provide neither error bars nor a description of the training-data distribution over Hamiltonian instances and hyper-parameters. This omission makes it impossible to assess the statistical reliability of the reported gains.

    Authors: We acknowledge that the abstract and main results section lack error bars and a clear description of the training-data distribution. The revised manuscript will include error bars obtained from five independent training runs with different random seeds. We will also expand the experimental setup to specify the distribution of TFIM instances (system sizes, transverse-field ranges) and the hyper-parameter sampling procedure used during training. revision: yes

Circularity Check

0 steps flagged

No significant circularity in claimed results or optimization chain

full rationale

The paper presents an empirical generative optimization pipeline (diffusion model proposals + PINN surrogate fidelity + GNN commutator encoding + REINFORCE loop) whose outputs are evaluated on TFIM instances. The headline metrics (85.6 % of Qiskit-4 fidelity at 21.8 % depth, 0.9994 under equal-depth fine-tuning) are stated as measured fidelities of the produced decompositions, not as quantities defined or fitted inside the surrogate itself. No equation, self-citation, or ansatz reduces the reported performance to a tautological re-expression of the training objective; the PINN is used only for differentiable guidance during search, while final numbers are presented as independent evaluations. No uniqueness theorem, renaming of known patterns, or load-bearing self-citation appears in the abstract or described method. The chain is therefore self-contained and non-circular.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that the PINN fidelity estimator is sufficiently accurate and differentiable to guide the diffusion model toward high-fidelity low-depth strategies, plus the implicit assumption that REINFORCE training on the hybrid discrete-continuous space converges reliably for the TFIM family.

free parameters (1)
  • CFG guidance scale
    Ablations indicate it must be tuned jointly with compute budget and affects module contributions.
axioms (1)
  • domain assumption The physics-informed neural network provides a faithful differentiable proxy for Trotter-Suzuki fidelity across the explored grouping and order space.
    Required for the generate-and-evaluate loop to produce meaningful gradients.

pith-pipeline@v0.9.0 · 5570 in / 1493 out tokens · 48696 ms · 2026-05-14T18:36:35.226274+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

35 extracted references · 21 canonical work pages · 2 internal anchors

  1. [1]

    Johnson, Jonathan Ho, Daniel Tarlow, and Rianne van den Berg

    Jacob Austin, Daniel D. Johnson, Jonathan Ho, Daniel Tarlow, and Rianne van den Berg. Structured denoising diffusion models in discrete state-spaces. InAdvances in Neural In- formation Processing Systems (NeurIPS), volume 34, pages 17981–17993, 2021

    2021

  2. [2]

    Berry, Andrew M

    Dominic W. Berry, Andrew M. Childs, Richard Cleve, Robin Kothari, and Rolando D. Somma. Simulating hamiltonian dynamics with a truncated taylor series.Physical Review Letters, 114(9):090502, 2015. doi: 10.1103/PhysRevLett.114.090502

    work page doi:10.1103/physrevlett.114.090502 2015

  3. [3]

    Kottmann, Tim Menke, Wai-Keong Mok, Sukin Sim, Leong-Chuan Kwek, and Al´ an Aspuru-Guzik

    Kishor Bharti, Alba Cervera-Lierta, Thi Ha Kyaw, Tobias Haug, Sumner Alperin-Lea, Abhinav Anand, Matthias Degroote, Hermanni Heimonen, Jakob S. Kottmann, Tim Menke, Wai-Keong Mok, Sukin Sim, Leong-Chuan Kwek, and Al´ an Aspuru-Guzik. Noisy intermediate-scale quantum algorithms.Reviews of Modern Physics, 94(1):015004, 2022. doi: 10.1103/RevModPhys.94.015004

    work page doi:10.1103/revmodphys.94.015004 2022

  4. [4]

    Theory of Trotter error with commutator scaling,

    Andrew M. Childs, Yuan Su, Minh C. Tran, Nathan Wiebe, and Shuchen Zhu. Theory of Trotter error with commutator scaling.Physical Review X, 11(1):011020, 2021. doi: 10.1103/PhysRevX.11.011020

    work page doi:10.1103/physrevx.11.011020 2021

  5. [5]

    Efficient quantum measurement of Pauli operators in the presence of finite sampling error.Quantum, 5:385, 2021

    Ophelia Crawford, Barnaby van Straaten, Daochen Wang, Thomas Parks, Earl Campbell, and Stephen Brierley. Efficient quantum measurement of Pauli operators in the presence of finite sampling error.Quantum, 5:385, 2021. doi: 10.22331/q-2021-01-20-385

    work page doi:10.22331/q-2021-01-20-385 2021

  6. [6]

    Scientific machine learning through physics-informed neural networks: where we are and what’s next.Journal of Scientific Computing, 92(3):88, 2022

    Salvatore Cuomo, Vincenzo Schiano Di Cola, Fabio Giampaolo, Gianluigi Rozza, Maziar Raissi, and Francesco Piccialli. Scientific machine learning through physics-informed neural networks: where we are and what’s next.Journal of Scientific Computing, 92(3):88, 2022. doi: 10.1007/s10915-022-01939-z

    work page doi:10.1007/s10915-022-01939-z 2022

  7. [7]

    Meyarivan

    Kalyanmoy Deb, Amrit Pratap, Sameer Agarwal, and T. Meyarivan. A fast and elitist mul- tiobjective genetic algorithm: NSGA-II.IEEE Transactions on Evolutionary Computation, 6(2):182–197, 2002. doi: 10.1109/4235.996017. 17

    work page doi:10.1109/4235.996017 2002

  8. [8]

    Phoenix: Pauli-based high-level opti- mization engine for instruction execution on NISQ devices.arXiv preprint, 2025

    Zeyuan Feng, Zhen Xie, Yufei Shi, and Gushu Li. Phoenix: Pauli-based high-level opti- mization engine for instruction execution on NISQ devices.arXiv preprint, 2025

    2025

  9. [9]

    Quantum circuit optimization with deep reinforcement learning.arXiv preprint arXiv:2103.07585, 2021

    Thomas F¨ osel, Murphy Yuezhen Niu, Florian Marquardt, and Li Li. Quantum circuit optimization with deep reinforcement learning.arXiv preprint arXiv:2103.07585, 2021

    work page arXiv 2021

  10. [10]

    Schoenholz, Patrick F

    Justin Gilmer, Samuel S. Schoenholz, Patrick F. Riley, Oriol Vinyals, and George E. Dahl. Neural message passing for quantum chemistry. InProceedings of the 34th International Conference on Machine Learning (ICML), pages 1263–1272, 2017

    2017

  11. [11]

    Pranav Gokhale, Olivia Angiuli, Yongshan Ding, Kaiwen Gui, Teague Tomesh, Martin Suchara, Margaret Martonosi, and Frederic T. Chong. Optimization of simultaneous mea- surement for variational quantum eigensolver applications. In2020 IEEE International Conference on Quantum Computing and Engineering (QCE), pages 379–390, 2020. doi: 10.1109/QCE49297.2020.00054

    work page doi:10.1109/qce49297.2020.00054 2020

  12. [12]

    Classifier-Free Diffusion Guidance

    Jonathan Ho and Tim Salimans. Classifier-free diffusion guidance.arXiv preprint arXiv:2207.12598, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  13. [13]

    Denoising diffusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising diffusion probabilistic models. In Advances in Neural Information Processing Systems (NeurIPS), volume 33, pages 6840– 6851, 2020

    2020

  14. [14]

    Equivariant diffusion for molecule generation in 3D

    Emiel Hoogeboom, V´ ıctor Garcia Satorras, Cl´ ement Vignac, and Max Welling. Equivariant diffusion for molecule generation in 3D. InProceedings of the 39th International Conference on Machine Learning (ICML), pages 8867–8887, 2022

    2022

  15. [15]

    Wood, Jake Lishman, Julien Gacon, Simon Martiel, Paul D

    Ali Javadi-Abhari, Matthew Treinish, Kevin Krsulich, Christopher J. Wood, Jake Lishman, Julien Gacon, Simon Martiel, Paul D. Nation, Lev S. Bishop, Andrew W. Cross, Blake R. Johnson, and Jay M. Gambetta. Quantum computing with Qiskit.arXiv preprint, 2024

    2024

  16. [16]

    Tetris: re- architecting convolutional neural network computation for machine learning accelerators

    Sung-Min Jin, Yi-Lin Liu, Yuan Su, Cheng Liang, Jian Tan, and Songhua Lu. Tetris: re- architecting convolutional neural network computation for machine learning accelerators. In Proceedings of the 50th Annual International Symposium on Computer Architecture (ISCA), pages 1–14, 2023

    2023

  17. [17]

    Kevrekidis, Lu Lu, Paris Perdikaris, Sifan Wang, and Liu Yang

    George Em Karniadakis, Ioannis G. Kevrekidis, Lu Lu, Paris Perdikaris, Sifan Wang, and Liu Yang. Physics-informed machine learning.Nature Reviews Physics, 3(6):422–440, 2021. doi: 10.1038/s42254-021-00314-5

    work page doi:10.1038/s42254-021-00314-5 2021

  18. [18]

    Paulihedral: a generalized block-wise com- piler optimization framework for quantum simulation kernels

    Gushu Li, Yufei Shi, and Ali Javadi-Abhari. Paulihedral: a generalized block-wise com- piler optimization framework for quantum simulation kernels. InProceedings of the 27th ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS), pages 554–569, 2022. doi: 10.1145/3503222.3507715

    work page doi:10.1145/3503222.3507715 2022

  19. [19]

    Lorenzo Moro, Matteo G. A. Paris, Marcello Restelli, and Enrico Prati. Quantum compiling by deep reinforcement learning.Communications Physics, 4(1):178, 2021. doi: 10.1038/ s42005-021-00684-3

    2021

  20. [20]

    Quantum Computing in the NISQ era and beyond.Quantum, 2:79, August 2018

    John Preskill. Quantum computing in the NISQ era and beyond.Quantum, 2:79, 2018. doi: 10.22331/q-2018-08-06-79

    work page internal anchor Pith review doi:10.22331/q-2018-08-06-79 2018

  21. [21]

    Maziar Raissi, Paris Perdikaris, and George Em Karniadakis. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations.Journal of Computational Physics, 378:686–707,

  22. [22]

    doi: 10.1016/j.jcp.2018.10.045. 18

    work page doi:10.1016/j.jcp.2018.10.045 2018

  23. [23]

    E(n) equivariant graph neural networks

    V´ ıctor Garcia Satorras, Emiel Hoogeboom, and Max Welling. E(n) equivariant graph neural networks. InProceedings of the 38th International Conference on Machine Learning (ICML), pages 9323–9332, 2021

    2021

  24. [24]

    Sch¨ utt, Huziel E

    Kristof T. Sch¨ utt, Huziel E. Sauceda, Pieter-Jan Kindermans, Alexandre Tkatchenko, and Klaus-Robert M¨ uller. SchNet: A continuous-filter convolutional neural network for modeling quantum interactions. InAdvances in Neural Information Processing Systems (NeurIPS), volume 30, 2017

    2017

  25. [25]

    Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole

    Yang Song, Jascha Sohl-Dickstein, Diederik P. Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. Score-based generative modeling through stochastic differential equations. InInternational Conference on Learning Representations (ICLR), 2021

    2021

  26. [26]

    General theory of fractal path integrals with applications to many-body theories and statistical physics,

    Masuo Suzuki. General theory of fractal path integrals with applications to many-body theories and statistical physics.Journal of Mathematical Physics, 32(2):400–407, 1991. doi: 10.1063/1.529425

    work page doi:10.1063/1.529425 1991

  27. [27]

    Fourier features let networks learn high frequency functions in low dimensional domains

    Matthew Tancik, Pratul Srinivasan, Ben Mildenhall, Sara Fridovich-Keil, Nithin Raghavan, Utkarsh Singhal, Ravi Ramamoorthi, Jonathan Barron, and Ren Ng. Fourier features let networks learn high frequency functions in low dimensional domains. InAdvances in Neural Information Processing Systems (NeurIPS), volume 33, pages 7537–7547, 2020

    2020

  28. [28]

    Tran, Yuan Su, Daniel Carney, and Jacob M

    Minh C. Tran, Yuan Su, Daniel Carney, and Jacob M. Taylor. Faster digital quantum simulation by symmetry protection.PRX Quantum, 2(1):010323, 2021. doi: 10.1103/ PRXQuantum.2.010323

    2021

  29. [29]

    Quantum graph neural networks.arXiv preprint arXiv:1909.12264,

    Guillaume Verdon, Trevor McCourt, Enxhell Luzhnica, Vikash Singh, Stefan Leichenauer, and Jack Hidary. Quantum graph neural networks.arXiv preprint arXiv:1909.12264, 2019

    work page arXiv 1909

  30. [30]

    Izmaylov

    Vladyslav Verteletskyi, Tzu-Ching Yen, and Artur F. Izmaylov. Measurement optimization in the variational quantum eigensolver using a minimum clique cover.The Journal of Chemical Physics, 152(12):124114, 2020. doi: 10.1063/1.5141458

    work page doi:10.1063/1.5141458 2020

  31. [31]

    When and why PINNs fail to train: A neural tangent kernel perspective.Journal of Computational Physics, 449:110768, 2022

    Sifan Wang, Xinling Yu, and Paris Perdikaris. When and why PINNs fail to train: A neural tangent kernel perspective.Journal of Computational Physics, 449:110768, 2022. doi: 10.1016/j.jcp.2021.110768

    work page doi:10.1016/j.jcp.2021.110768 2022

  32. [32]

    A fast way of calculating exact hy- pervolumes.IEEE Transactions on Evolutionary Computation, 16(1):86–95, 2012

    Lyndon While, Lucas Bradstreet, and Luigi Barone. A fast way of calculating exact hy- pervolumes.IEEE Transactions on Evolutionary Computation, 16(1):86–95, 2012. doi: 10.1109/TEVC.2010.2077298

    work page doi:10.1109/tevc.2010.2077298 2012

  33. [33]

    Berry, Peter Høyer, and Barry C

    Nathan Wiebe, Dominic W. Berry, Peter Høyer, and Barry C. Sanders. Higher order decompositions of ordered operator exponentials.Journal of Physics A: Mathematical and Theoretical, 43(6):065203, 2010. doi: 10.1088/1751-8113/43/6/065203

    work page doi:10.1088/1751-8113/43/6/065203 2010

  34. [34]

    , title =

    Ronald J. Williams. Simple statistical gradient-following algorithms for connectionist rein- forcement learning.Machine Learning, 8:229–256, 1992. doi: 10.1007/BF00992696

    work page doi:10.1007/bf00992696 1992

  35. [35]

    Fonseca, and Viviane Grunert da Fonseca

    Eckart Zitzler, Lothar Thiele, Marco Laumanns, Carlos M. Fonseca, and Viviane Grunert da Fonseca. Performance assessment of multiobjective optimizers: an analysis and review. IEEE Transactions on Evolutionary Computation, 7(2):117–132, 2003. doi: 10.1109/TEVC. 2003.810758. 19

    work page doi:10.1109/tevc 2003