Pith. sign in

REVIEW 3 major objections 6 minor 20 references

Empowering Neural Network-based Quantum Monte Carlo with Local Pseudopotentials

T0 review · 3 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read A fully local pseudo-Hamiltonian, not all-electron or semilocal effective-core forms, is the efficient, accurate Hamiltonian choice for neural-network quantum Monte Carlo, taking it to 268-electron iron-sulfur clusters.

desk verdict Promising integration of local pseudo-Hamiltonians with NNQMC, but the Forward Laplacian derivation in Eq. (15) wrongly treats the position-dependent Q_i as constant, undercutting the reported efficiency and accuracy claims until fixed. read the letter →

arxiv 2505.19909 v2 pith:ZW64MKX5 submitted 2025-05-26 physics.chem-ph

classification physics.chem-ph
keywords neuralnetworkquantumMonteCarlolocalpseudopotentialpseudo-Hamiltonianeffectivecorepotentialiron-sulfurclusterForwardLaplacianvariationaldiffusion
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

This paper argues that neural-network quantum Monte Carlo (NNQMC) becomes far more practical when the Hamiltonian is a local pseudo-Hamiltonian (PH) rather than the all-electron Coulomb form or the standard semilocal effective-core potential. The authors construct a PH for sulfur by fitting it to ccECP reference data, and combine it with the Forward Laplacian derivative scheme so that the PH's angular-momentum-squared term adds almost no cost. They report over tenfold faster training, accuracy equal to or better than all-electron and ECP calculations, and the first NNQMC treatment of a 268-electron iron-sulfur cubane cluster. If correct, this removes the main scaling bottleneck of NNQMC and extends the method to transition-metal chemistry.

What carries the argument

The load-bearing object is the pseudo-Hamiltonian $h^{\mathrm{PH}}(r) = -\tfrac{1}{2}\, \hat{p}\, a(r)\, \hat{p} + V^{\mathrm{PH}}_{\mathrm{local}}(r) + V_{L2}(r)\, \hat{L}^2$, with $a(r)=0$ in this work so the kinetic operator keeps its standard form. Unlike a semilocal effective core potential, this form has no $|\ell m\rangle\langle\ell m|$ projectors; the angular dependence is carried by $\hat{L}^2$, and inside the Forward Laplacian scheme the required second derivatives are evaluated through a per-electron coordinate scaling $v_i = Q_i^{-1} r_i$, turning them into ordinary Laplacians with negligible extra cost. For sulfur, the potential functions are fitted so that their channel matrix elements reproduce the ccECP reference values (Eq. 11), under the linear-dependency constraint $2V_0(r)-3V_1(r)=0$ and the positivity bound $1+2r^2 V_{L2}(r)>0$, using atomic eigenenergies, norm-conservation outside the cutoff, excitation energies, and the $S_2$/FeS binding curves as reference data. This single construction carries both the efficiency gain (fewer electrons, no spherical-harmonic integrals) and the accuracy gain (smooth core region without nuclear Coulomb singularities).

What would settle it

Take the fitted sulfur pseudo-Hamiltonian and compute the dissociation curve of a sulfur-containing system outside the fitting set, for instance SO$_2$ at a highly stretched geometry or a periodic sulfur crystal, using NNQMC with PH, and compare against high-level CCSD(T) at the complete-basis limit. A systematic error growing beyond chemical accuracy with geometry would show that the ccECP-fitted local potential does not transfer beyond its training data.

Watch

Extended reading notes

Core claim

The paper's central claim is that the pseudo-Hamiltonian (PH), a fully local pseudopotential that replaces the nonlocal spherical-harmonic projectors of an effective core potential with the angular-momentum-squared operator $\hat{L}^2$ and a radial effective-mass term, is the preferred Hamiltonian choice inside NNQMC. Setting $a(r)=0$ keeps the kinetic term standard, and the $\hat{L}^2$ term is absorbed into a Forward Laplacian coordinate transformation so its second derivatives are nearly free. Across sulfur chains, sulfur-containing molecules, transition-metal sulfides, and two iron-sulfur clusters, the authors find PH over tenfold faster than both all-electron and ECP treatments, accuracy matching or beating all-electron results, and agreement with ECP, MRCC, and experiment. They further obtain a magnetic exchange coupling constant $J=221\,\mathrm{cm}^{-1}$ for $[\mathrm{Fe}_2\mathrm{S}_2(\mathrm{SCH}_3)_4]^{2-}$, consistent with experiments and high-level theory, and a physically meaningful spin distribution for the cubane cluster $[\mathrm{Fe}_4\mathrm{S}_4(\mathrm{SCH}_3)_4]$. The paper concludes that PH generally enhances NNQMC's computational efficiency by over tenfold and enables reliable treatment of systems that were previously beyond the method's reach.

Load-bearing premise

The sulfur pseudo-Hamiltonian is fitted to ccECP reference data for a few atomic and dimer properties, and the paper's broad accuracy claims depend on that fitted potential transferring reliably to every sulfur environment, molecule, and NNQMC calculation tested; if the fit only works near its training set, the general accuracy result does not follow.

Editorial extensions

If this is right

  • NNQMC can now handle transition-metal clusters with hundreds of electrons, such as the 268-electron cubane $[\mathrm{Fe}_4\mathrm{S}_4(\mathrm{SCH}_3)_4]$, with over twentyfold acceleration.
  • The pseudo-Hamiltonian is strictly more efficient than semilocal ECPs inside NNQMC, because the $\hat{L}^2$ term costs almost nothing under the Forward Laplacian scheme while ECP's spherical-harmonic integrals dominate the runtime.
  • Removing core electrons improves convergence as well as cost: PH reaches chemical accuracy after about 50,000 training iterations, whereas all-electron training is still far from convergence after 150,000.
  • PH-based NNQMC yields quantitatively reliable magnetic observables, including $J=221\,\mathrm{cm}^{-1}$ for the dinuclear iron-sulfur cluster, positioning the method as a practical tool for strongly correlated bioinorganic systems.

Reading between the lines

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

  • If the same fitting recipe transfers element by element, a library of local pseudo-Hamiltonians would make NNQMC routine for transition-metal and period-3 chemistry, not just the sulfur and iron cases demonstrated here.
  • The Forward-Laplacian-plus-$\hat{L}^2$ trick should carry over to periodic solids through solid-state neural ansatz, potentially opening d- and f-electron solid-state problems to NNQMC.
  • A decisive test is systematic transferability benchmarking across oxidation states and coordination environments: the paper notes that PH construction still requires system-specific tuning, so the practical payoff depends on how much re-fitting each new element needs.
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 / 6 minor

Summary. The paper proposes to use local pseudo-Hamiltonians (PHs) inside neural-network quantum Monte Carlo (NNQMC) as a cheaper and more accurate alternative to all-electron (AE) and semilocal effective-core-potential (ECP) Hamiltonians. The authors construct a sulfur PH by fitting to ccECP reference data, combine it with existing transition-metal PHs, and report over-tenfold speedups, near-chemical-accuracy benchmarks on sulfur molecules and transition-metal sulfides, and a demonstration on Fe2S2(SCH3)4 and Fe4S4(SCH3)4. The core technical claim is that the PH's L^2 term can be evaluated at almost no extra cost through a Forward-Laplacian coordinate transformation.

Significance. If the implementation is correct, the manuscript would address a real bottleneck: nonlocal ECP projections are expensive in NNQMC, and local PHs plus the Forward Laplacian could substantially extend NNQMC to transition-metal systems. The paper has several genuine strengths: the PH construction procedure in Section 4.3 and Supplementary Note 4 is detailed, the code is open-sourced, and the iron-sulfur cluster applications are ambitious. However, the central efficiency derivation in Section 4.4 appears mathematically inconsistent as written, and the accuracy claims in Fig. 3 are partly tested against the same ccECP data used to fit the sulfur PH. The promise of the approach is clear, but the current manuscript does not yet establish the stated Hamiltonian, speedups, or transferability at the claimed level.

major comments (3)
  1. The Forward-Laplacian identity is not valid as written for the PH operator. Equation (14) defines v_i = Q_i^{-1} r_i with Q_i an 'invertible constant matrix' for the decomposition A(r_i) = Q_i^T Q_i, but A_{alpha beta}(r_i) in Supplementary Eq. (22) depends on the electron position through V_{L2,I}(|r_i - R_I|). A position-dependent A cannot be represented by a single constant Q_i per electron and geometry. The exact chain rule for an r_i-dependent Q_i contains additional first-order derivative terms involving partial derivatives of Q_i, and these terms are absent from Eq. (15) and from the explicit b_alpha formula in Supplementary Eq. (23). If the implementation evaluates the Laplacian by treating Q_i as a detached constant, then the local energy computed is not the pseudo-Hamiltonian of Eq. (8). This issue sits under both the claimed 'negligible overhead' for the L^2 term and every PH energy reported in Figs. 3-4. Please provide the corrected identities including the Q_i-derivative contributions, or state explicitly which operator was actually implemented and demonstrate that the omitted terms vanish by construction or are negligible.
  2. The sulfur PH is fitted directly to ccECP reference data: atomic eigenenergies, norm conservation, excitation energies, the S2 binding curve, and the FeS binding curve (cost function in Supplementary Note 4, Eqs. (4)-(13), and step 2 of Section 4.3). The close PH/ECP agreement for the S2 potential energy curve in Fig. 3c and for FeS in Fig. 3d is therefore partially a consistency check of the fitting procedure rather than an independent validation. To support the transferability claim, the authors should explicitly separate training systems from validation systems, or benchmark the sulfur PH on sulfur environments (and molecules) that were not used in the optimization, and then reassess the wording of the accuracy conclusions.
  3. No statistical uncertainties are reported for the energy differences in Fig. 3 or for the timing comparisons in Fig. 2. NNQMC energies are stochastic; the text reports values like -467.8407(3) in Supplementary Table 2, so error bars are available for some quantities, but the central benchmark plots omit them. Statements such as 'PH achieves chemical accuracy for most species' (Fig. 3b), 'within the experimental uncertainties' (Fig. 3d), and 'over tenfold' acceleration (Section 2.2) cannot be assessed without error bars or repeated-run estimates. Please add uncertainties to all plotted quantities and to the efficiency ratios, and qualify conclusions accordingly.
minor comments (6)
  1. Typo: 'psedo-Hamiltonian' should be 'pseudo-Hamiltonian'.
  2. Typo: 'Dissociatino Energy Error' should be 'Dissociation Energy Error'.
  3. The symbol i is used both as an electron index and as the imaginary unit; this creates confusion. Use \mathrm{i} for the imaginary unit or rename the electron index.
  4. The inset magnifying the PH and ECP convergence curves is too small to read; please enlarge it or show the same data in a separate panel.
  5. The statement that the coordinate transformation 'can be easily integrated with Forward Laplacian by modifying the input' is too brief for a reader wishing to reproduce the implementation; give explicit formulas for how the modified input and the derivative pattern are constructed in the code.
  6. The magnetic exchange coupling constant J is quoted without an uncertainty. Since both E_BS, E_HS and the spin-squared expectation values are stochastic, please quote a statistical error for J.

Circularity Check

3 steps flagged · score 6.0 of 10

Several sulfur-PH 'validation' benchmarks are training targets; the central NNQMC+PH efficiency and application claims remain independently supported.

  1. fitted input called prediction [Section 2.3 / Fig. 3a; Section 4.3 step 1; Supplementary Note 4 Eqs. (5)-(7)]
    "“As a first validation, we benchmark PH on spin gap, IE, and EA of the sulfur atom against experimental references [42–44]. As shown in Fig. 3a, PH shows excellent agreement with ECP and achieves near-exact accuracy for the spin gap and EA relative to experiment.” The cost-function description states: “For the excited states, high-spin states ranging from charge -1 to +5, as well as any lower spin states (if any), were considered.”"

    The sulfur PH is fitted by minimizing O_eigenenergy, O_norm, and O_excitation deviations from ccECP for the ground configuration 3s2 3p4 and for charge states from -1 to +5 (Supplementary Note 4). The spin gap, ionization energy, and electron affinity displayed in Fig. 3a are energy differences among exactly those fitted atomic states. Reporting 'near-exact accuracy' for them is therefore reporting the outcome of the fit against the ccECP/CCSD(T) reference used to construct the PH; the experimental comparison is informative, but the PH/ECP agreement in the figure is by construction. The paper itself notes that the IE residual reflects the CCSD(T) reference used during construction.

  2. fitted input called prediction [Section 2.3 / Fig. 3c; Section 4.3 steps 2-5; Supplementary Note 4 Eqs. (4), (8), (11)]
    "“The term O^HF_S2binding decreases as the binding curve of the reference S2 dimer is better reproduced.” In the main text: “we generate the potential energy curve (PEC) of the sulfur dimer ... The PH and ECP PECs align closely, both exhibiting low NPEs (respectively 2.4 and 2.0 mHa).”"

    The S2 dimer binding curve is an explicit training target twice over: O^HF_S2binding enters the cost function, and steps 2-3 use Bayesian optimization to minimize the CCSD(T) mean absolute error of the S2 and FeS binding curves, followed by re-optimization at approximate CCSD(T) accuracy in step 5. The low-NPE PEC in Fig. 3c is thus selected, not predicted, from the reference S2 curve (ccECP/CCSD(T)); only the MRCI+Q-F12 comparison provides independent information, and even that comparison is partly forced to the extent CCSD(T) and MRCI+Q-F12 agree on S2.

1 more flagged steps
  1. fitted input called prediction [Section 4.3 steps 2-3; Fig. 3d (FeS point)]
    "“Based on the optimized pseudo-Hamiltonian, evaluate the binding curve of S2 and FeS dimers, and compute the mean absolute error (MAE) with respect to the ccECP results using the CCSD(T) method. Update the weights of the cost function by Bayesian optimization to minimize the MAE values obtained in step 2.”"

    The FeS dissociation curve is one of the two molecular binding curves directly used in the sulfur-PH cost function: its CCSD(T) MAE vs ccECP is the quantity minimized by Bayesian weight updates (Section 4.3 steps 2-3) and by the approximate-CCSD(T) re-optimization (step 5). Presenting FeS among the 'transition metal sulfides' dissociation-energy benchmarks in Fig. 3d therefore reports a training target as a validation; its close agreement is partly by construction. The MnS, CoS, CuS, and ZnS points are not in the sulfur-PH training set and remain genuine transfer predictions.

full rationale

Three Fig. 3 benchmarks are training targets of the newly constructed sulfur pseudo-Hamiltonian: S-atom spin gap/IE/EA, the S2 PEC, and the FeS dissociation energy. These are not independent confirmations, although the paper is transparent that the PH is fitted to ccECP atomic and S2/FeS reference data. The central contribution—that a fully local pseudo-Hamiltonian removes ECP's nonlocal spherical-harmonic cost under the Forward Laplacian and makes NNQMC practical for iron-sulfur clusters—does not reduce to those fits: atomization energies of other sulfur molecules and MnS/CoS/CuS/ZnS are novel transfer tests, and the [Fe2S2(SCH3)4]2- and [Fe4S4(SCH3)4] simulations are new applications benchmarked against DMRG, CCSD(T), MRCC, and experiment. The transition-metal PHs are imported from prior same-author papers [29,30], but those are separately published and are benchmarked here against external references, so the self-citation is not load-bearing. The Section 4.4 Forward-Laplacian chain-rule identity (Eq. 15) is a mathematical-correctness concern involving the position dependence of Q_i and would matter for the efficiency/energy claims, but it is not a reduction of a prediction to its input, so it does not by itself contribute to the circularity score.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central calculation rests on (i) the PH parameterization for sulfur, whose parameters and cost weights are fitted to ccECP reference data, and (ii) standard QMC and pseudopotential assumptions: Born-Oppenheimer, fixed-node, core freezing, and transferability. No new physical entities are postulated. The fitting of PH to the same reference used in some benchmarks is the main source of circularity burden.

free parameters (3)
  • Sulfur pseudo-Hamiltonian potential parameters (V_PH_local and V_L2 Gaussian coefficients) = Not reported in main text; optimized numerically and stored in the JaQMC repository.
    Section 4.3 and Supplementary Note 4: parameters are optimized so that PH matrix elements match ccECP (Eq. 11) on atomic eigenenergies, norms, excitations, and S2/FeS binding curves. These are fitted to reference data, not derived.
  • Cost-function weights omega_1 to omega_4 and lambda = Not reported; set by Bayesian optimization over 50 iterations.
    Supplementary Note 4: weights are updated to minimize the mean absolute error of S2 and FeS binding curves versus ccECP, so they are fitted parameters that shape the final PH.
  • V'_L2 correction-term Gaussian coefficients = Not reported; optimizable coefficients with exponents fixed at 3.0, 5.0, and 7.0 Bohr^-2.
    Added to satisfy the bounding condition 1 + 2 r^2 V_L2 > 0; the coefficients are tuned during optimization.
assumptions (5)
  • domain assumption Born-Oppenheimer separation of electronic and nuclear motion.
    Section 4.2 constructs Hamiltonians under the Born-Oppenheimer approximation, standard for molecular electronic structure but still an assumption.
  • domain assumption Fixed-node approximation in DMC yields the fermionic ground state without the sign problem.
    Section 4.1 relies on the fixed-node approximation to prevent the sign problem; its quality depends on the trial nodal surface.
  • domain assumption Core-valence separation: chemically inert core electrons can be frozen and replaced by pseudopotentials without loss of chemical accuracy.
    Sections 2.1 and 4.2 replace all-electron Coulomb cores with PH/ECP, the standard pseudopotential assumption.
  • ad hoc to paper The local PH form with a(r)=0 and L^2 term, plus constraints 2V0 - 3V1 = 0 and 1 + 2r^2 V_L2 > 0, can faithfully represent the ccECP for sulfur.
    Eq. (11) and Supplementary Note 4 enforce a linear dependency and a bounding condition so PH matches ECP matrix elements; this is an approximation specific to the construction.
  • domain assumption The sulfur PH fitted to ccECP atomic and dimer data transfers to other molecules and to NNQMC with LapNet.
    Section 4.3 optimizes PH against atomic eigenenergies, norms, excitations, and S2/FeS binding curves, then applies it to a broader benchmark set; transferability is assumed rather than proven.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Empowering Neural Network-based Quantum Monte Carlo with Local Pseudopotentials." pith.science (2026). https://pith.science/paper/ZW64MKX5

@misc{pith2026250519909,
  author       = {Pith},
  title        = {Pith review of: Empowering Neural Network-based Quantum Monte Carlo with Local Pseudopotentials},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZW64MKX5}},
  note         = {Machine review of arXiv:2505.19909}
}
abstract

Neural Network-based Quantum Monte Carlo (NNQMC), an emerging method for solving many-body quantum systems with high accuracy, has been mainly applied to small systems due to demanding computation requirements. In this work, we introduce a framework based on local pseudopotentials to break through such limitation, improving the computational efficiency and scalability of NNQMC. The incorporation of local pseudopotentials reduces the number of electrons treated in neural network and also achieves better relative energy accuracy than all electron NNQMC calculations for complex systems. This counterintuitive outcome is made possible by the distinctive characteristics inherent to NNQMC. Notably, by avoiding costly integration terms, this approach is also substantially more efficient than its widely used semilocal counterparts. Our approach enables the reliable treatment of large and challenging systems, such as the $\text{Fe}_4 \text{S}_4 (\text{SCH}_3)_4$ iron-sulfur cluster. Overall, our findings demonstrate that the synergy between NNQMC and local pseudopotentials substantially expands the scope of accurate ab initio calculations.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages

  1. [1]

    Energy-consistentpseudopotentialsforquantummontecarlocalculations

    MarkBurkatzki,CFilippi,andMDolg. Energy-consistentpseudopotentialsforquantummontecarlocalculations. The Journal of chemical physics, 126(23), 2007

  2. [2]

    Chandler Bennett, Guangming Wang, Abdulgani Annaberdiyev, Cody A

    M. Chandler Bennett, Guangming Wang, Abdulgani Annaberdiyev, Cody A. Melton, Luke Shulenburger, and Lubos Mitas. A new generation of effective core potentials from correlated calculations: 2nd row elements.J. Chem. Phys, 149:104108, Sep 2018

  3. [3]

    Novel pseudo-hamiltonian for quantum monte carlo simulations.Physical review letters, 62(18):2088, 1989

    GB Bachelet, DM Ceperley, and MGB Chiocchetti. Novel pseudo-hamiltonian for quantum monte carlo simulations.Physical review letters, 62(18):2088, 1989. 6

  4. [4]

    Localityerrorfreeeffectivecorepotentialsfor3dtransitionmetalelementsdevelopedforthediffusionmontecarlomethod

    Tom Ichibha, Yutaka Nikaido, M Chandler Bennett, Jaron T Krogel, Kenta Hongo, Ryo Maezono, and Fernando A Reboredo. Localityerrorfreeeffectivecorepotentialsfor3dtransitionmetalelementsdevelopedforthediffusionmontecarlomethod. The Journal of Chemical Physics, 159(16), 2023

  5. [5]

    High accuracy transition metal effective cores for the many-body diffusion monte carlo method.Journal of Chemical Theory and Computation , 18(2):828–839, 2022

    M Chandler Bennett, Fernando A Reboredo, Lubos Mitas, and Jaron T Krogel. High accuracy transition metal effective cores for the many-body diffusion monte carlo method.Journal of Chemical Theory and Computation , 18(2):828–839, 2022

  6. [6]

    Hybridizingpseudo-hamiltoniansandnon-localpseudopotentialsindiffusionmonte carlo

    JaronTKrogelandFernandoAReboredo. Hybridizingpseudo-hamiltoniansandnon-localpseudopotentialsindiffusionmonte carlo. The Journal of Chemical Physics, 153(10), 2020

  7. [7]

    Levine, Eric L

    Svetlana Kotochigova, Zachary H. Levine, Eric L. Shirley, M. D. Stiles, and Charles W. Clark. Local-density-functional calcu- lations of the energy of atoms.Phys. Rev. A, 55:191–199, Jan 1997

  8. [8]

    Levine, Eric L

    Svetlana Kotochigova, Zachary H. Levine, Eric L. Shirley, M. D. Stiles, and Charles W. Clark. Erratum: Local-density- functional calculations of the energy of atoms [Phys. Rev. A 55, 191 (1997)].Phys. Rev. A, 56:5191–5192, Dec 1997

Show all 20 references
  1. [9]

    Knowles, Gerald Knizia, Frederick R

    Hans-Joachim Werner, Peter J. Knowles, Gerald Knizia, Frederick R. Manby, and Martin Schütz. Molpro: a general-purpose quantum chemistry program package.Wiley Interdiscip. Rev.: Comput. Mol. Sci., 2(2):242–253, July 2011

  2. [10]

    Knowles, Frederick R

    Hans-Joachim Werner, Peter J. Knowles, Frederick R. Manby, Joshua A. Black, Klaus Doll, Andreas Heßelmann, Daniel Kats, Andreas Köhn, Tatiana Korona, David A. Kreplin, Qianli Ma, Thomas F. Miller, Alexander Mitrushchenkov, Kirk A. Peterson, Iakov Polyak, Guntram Rauhut, and Ma...

  3. [11]

    Peterson and Jr

    Kirk A. Peterson and Jr. Dunning, Thom H. Accurate correlation consistent basis sets for molecular core–valence correlation effects: The second row atoms al–ar, and the first row atoms b–ne revisited.J. Chem. Phys., 117(23):10548, Dec 2002

  4. [12]

    A new generation of effective core potentials from correlated calculations: 3d transition metal series.The Journal of chemical physics, 149(13), 2018

    Abdulgani Annaberdiyev, Guangming Wang, Cody A Melton, M Chandler Bennett, Luke Shulenburger, and Lubos Mitas. A new generation of effective core potentials from correlated calculations: 3d transition metal series.The Journal of chemical physics, 149(13), 2018

  5. [13]

    Spendley, G

    W. Spendley, G. R. Hext, and F. R. Himsworth. Sequential Application of Simplex Designs in Optimisation and Evolutionary Operation. Technometrics, 4(4):441–461, 1962

  6. [14]

    J. A. Nelder and R. Mead. A Simplex Method for Function Minimization.Comput. J., 7(4):308–313, 01 1965

  7. [15]

    Oliphant, Matt Haberland, Tyler Reddy, David Cournapeau, Evgeni Burovski, Pearu Peterson, WarrenWeckesser, JonathanBright, StéfanJ

    Pauli Virtanen, Ralf Gommers, Travis E. Oliphant, Matt Haberland, Tyler Reddy, David Cournapeau, Evgeni Burovski, Pearu Peterson, WarrenWeckesser, JonathanBright, StéfanJ. vanderWalt, Matthew Brett, Joshua Wilson, K.JarrodMillman, Niko- lay Mayorov, Andrew R. J. Nelson, Eric J...

  8. [16]

    Hyperopt: a python library for model selection and hyperparameter optimization.Comput

    James Bergstra, Brent Komer, Chris Eliasmith, Dan Yamins, and David D Cox. Hyperopt: a python library for model selection and hyperparameter optimization.Comput. Sci. Discov., 8:014008, Jul 2015

  9. [17]

    Algorithms for hyper-parameter optimization

    James Bergstra, Rémi Bardenet, Yoshua Bengio, and Balázs Kégl. Algorithms for hyper-parameter optimization. InAdvances in Neural Information Processing Systems, volume 24. Curran Associates, Inc., 2011

  10. [18]

    Jaron T. Krogel. Nexus: A modular workflow management system for quantum simulation codes.Comput. Phys. Commun., 198:154 – 168, 2016

  11. [19]

    Pyqmc: An all-pythonreal-space quantum monte carlomodule in pyscf

    William A Wheeler, Shivesh Pathak, Kevin G Kleiner, Shunyue Yuan, João NB Rodrigues, Cooper Lorsung, Kittithat Krong- chon, Yueqing Chang, Yiqing Zhou, Brian Busemeyer, et al. Pyqmc: An all-pythonreal-space quantum monte carlomodule in pyscf. The Journal of Chemical Physics, 1...

  12. [20]

    Low-energy spectrum of iron–sulfur clusters directly from many-particle quantum mechanics.Nature chemistry, 6(10):927–933, 2014

    Sandeep Sharma, Kantharuban Sivalingam, Frank Neese, and Garnet Kin-Lic Chan. Low-energy spectrum of iron–sulfur clusters directly from many-particle quantum mechanics.Nature chemistry, 6(10):927–933, 2014. 7

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.