pith. sign in

arxiv: 2603.12203 · v4 · pith:RMLGSMS4new · submitted 2026-03-12 · 🪐 quant-ph

Sparse Phase Ansatzes for Resource-Efficient Qudit State Preparation via the SNAP-Displacement Protocol

Maurizio Ferrari Dacrema , Doga Murat Kurkcuoglu , Andy C. Y. Li , Tanay Roy , Silvia Zorzetti This is my paper

Pith reviewed 2026-05-21 10:54 UTC · model grok-4.3

classification 🪐 quant-ph
keywords sparse ansatzesSNAP-displacement protocolqudit state preparationresource-efficient quantum controlmulti-objective optimizationbosonic statesfidelity-resource trade-offsnoisy quantum devices
0
0 comments X

The pith

Sparse subsets of SNAP phases in the displacement protocol deliver better fidelity-resource trade-offs for bosonic qudit state preparation than full parameterization.

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

The paper introduces three progressively more general sparse ansatzes that optimize only a chosen subset of SNAP phases within the SNAP-displacement protocol instead of tuning every phase. These ansatzes are paired with a scalarized multi-objective optimizer that explicitly balances achieved fidelity against either the total number of phases or the overall pulse duration. Numerical tests on multiple target states and qudit sizes up to dimension 64 show that the resulting Pareto fronts improve on the fully parameterized protocol in both ideal simulations and under photon-loss and control-error models. The gains are largest and most consistent when the objective is to minimize phase count; gains in duration are smaller and vary with the state family and noise level.

Core claim

By restricting optimization to subsets of SNAP phases via three sparse ansatzes and trading fidelity against resource counts through multi-objective optimization, the SNAP-displacement protocol reaches competitive state-preparation fidelities with substantially lower gate counts and shorter ansatz durations than the fully parameterized version, as measured by hypervolume of the Pareto frontiers for dimensions up to 64 in both ideal and noisy regimes.

What carries the argument

Three sparse ansatzes that select and optimize only a subset of the SNAP phases inside the SNAP-displacement protocol, combined with scalarized multi-objective optimization that trades fidelity against phase count or ansatz duration.

If this is right

  • The largest and most reliable resource savings occur when the objective is to reduce the number of SNAP phases rather than ansatz duration.
  • Performance advantages persist under realistic photon-loss and control-error noise, making the approach suitable for near-term hardware.
  • Gains in duration are smaller and more dependent on the specific target-state family.
  • The method supplies a tunable knob for selecting the ansatz that best matches a required fidelity on a given device.

Where Pith is reading between the lines

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

  • The same sparsity idea could be tested on other phase-gate families beyond SNAP to see whether similar resource reductions appear.
  • Hardware runs with calibrated pulse shapes would reveal whether the simulated Pareto improvements survive real control imperfections.
  • Extending the ansatzes to still higher dimensions or to continuous-variable states would test how far the subset-optimization pattern generalizes.

Load-bearing premise

Optimizing only a chosen subset of SNAP phases remains sufficient to reach competitive fidelity for the target states and noise models examined.

What would settle it

An experiment or simulation on a new target state or noise model in which the best sparse ansatz cannot match the fidelity of the fully parameterized protocol even after exhaustive optimization of the chosen phases.

Figures

Figures reproduced from arXiv: 2603.12203 by Andy C. Y. Li, Doga Murat Kurkcuoglu, Maurizio Ferrari Dacrema, Silvia Zorzetti, Tanay Roy.

Figure 3
Figure 3. Figure 3: FIG. 3: Infidelity and number of non-zero phase angles for the best hyperparameter configuration [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5: Pareto frontiers showing the trade-off between the infidelity of the prepared state and the number of [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6: Infidelity and duration of the ansatz for the best hyperparameter configuration [PITH_FULL_IMAGE:figures/full_fig_p019_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7: Infidelity and number of non-zero phase angles for the best hyperparameter configuration [PITH_FULL_IMAGE:figures/full_fig_p020_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8: Pareto frontiers showing the trade-off between the infidelity of the prepared state and the number of [PITH_FULL_IMAGE:figures/full_fig_p024_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9: Pareto frontiers showing the trade-off between the infidelity of the prepared state and the duration of the [PITH_FULL_IMAGE:figures/full_fig_p028_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10: Pareto frontiers showing the trade-off between the infidelity of the prepared state and the number of [PITH_FULL_IMAGE:figures/full_fig_p030_10.png] view at source ↗
read the original abstract

Efficient preparation of nonclassical bosonic states is a central requirement for quantum computing, simulation, and precision metrology. We study resource-efficient quantum state preparation in bosonic qudit systems using the SNAP-displacement (SD) protocol. Existing SD-based approaches typically require a large number of gates and SNAP phases, resulting in complex control pulses with longer ansatz durations and amplified impact of photon-loss and control errors. In this work, we focus on the near- to medium-term regime, in which noisy quantum devices impose trade-offs on the fidelity that can be achieved, which must be taken into account. Specifically, we propose to optimize only a subset of the SNAP phases and introduce three progressively more general sparse ansatzes. To provide fine-grained control and identify the most suitable ansatz for a given target fidelity, we further employ a scalarized multi-objective optimization that trades off fidelity against either the number of phases or the duration of the ansatz. Numerical results for several target states and qudit dimensions up to $d=64$, evaluated through the hypervolume of the Pareto frontiers, show that these sparse ansatzes achieve favorable trade-offs over the fully parameterized SD protocol in both ideal and noisy settings. The advantage is strongest and most consistent when minimizing the number of phases, while improvements in ansatz duration are smaller and more dependent on the target-state family and noise level, suggesting a practical route to more efficient near- and medium-term bosonic state preparation.

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

1 major / 2 minor

Summary. The manuscript proposes three sparse ansatzes that optimize only subsets of SNAP phases within the SNAP-displacement protocol for bosonic qudit state preparation. It employs scalarized multi-objective optimization to balance fidelity against either the number of phases or ansatz duration, then evaluates the resulting Pareto fronts via hypervolume for several target states and dimensions up to d=64. Numerical results are reported to show that the sparse ansatzes achieve superior trade-offs relative to the fully parameterized SD protocol in both ideal and noisy regimes, with the strongest gains when minimizing phase count.

Significance. If the numerical comparisons prove robust under equivalent optimization effort, the work would supply a concrete method for lowering control complexity and error sensitivity in near-term bosonic devices. The inclusion of noise models and demonstrations at d=64 add practical value; the hypervolume metric supplies a clear quantitative basis for comparing resource-fidelity frontiers. The central claim remains defensible but requires clarification on optimization fairness to reach full impact.

major comments (1)
  1. [Numerical results] Numerical results section: the headline claim that sparse ansatzes yield favorable hypervolume trade-offs rests on direct comparison of Pareto fronts obtained from the three proposed subsets versus the full O(d)-parameter SD protocol. The manuscript provides no information on the optimizer (e.g., evolutionary algorithm, gradient method), population size, iteration budget, or whether search effort was scaled with parameter-space dimension. Because the full protocol operates in a higher-dimensional space, any fixed-budget search could converge more readily for the sparse cases, potentially inflating the reported advantage. This is load-bearing for the central representational claim and must be addressed with explicit controls or re-optimization under matched effort.
minor comments (2)
  1. [Abstract] Abstract: the three sparse ansatzes are described only as 'progressively more general' without even a one-sentence characterization of their sparsity patterns; a brief parenthetical definition would aid immediate comprehension.
  2. [Methods] Methods or supplementary material: the precise noise model (photon-loss rates, control-error channels, and how they enter the fidelity objective) is not stated in the provided text; explicit equations or parameter values are needed for reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their thorough review and constructive feedback on our manuscript. The major comment identifies a valid and important issue regarding the transparency and fairness of the optimization comparisons. We have revised the manuscript to supply the missing details on the optimizer and to incorporate explicit controls that equalize search effort across ansatzes of differing dimensionality. These additions improve the robustness and reproducibility of the reported results.

read point-by-point responses

  1. Referee: [Numerical results] Numerical results section: the headline claim that sparse ansatzes yield favorable hypervolume trade-offs rests on direct comparison of Pareto fronts obtained from the three proposed subsets versus the full O(d)-parameter SD protocol. The manuscript provides no information on the optimizer (e.g., evolutionary algorithm, gradient method), population size, iteration budget, or whether search effort was scaled with parameter-space dimension. Because the full protocol operates in a higher-dimensional space, any fixed-budget search could converge more readily for the sparse cases, potentially inflating the reported advantage. This is load-bearing for the central representational claim and must be addressed with explicit controls or re-optimization under matched effort.

    Authors: We agree that the absence of optimizer details and effort-scaling information is an oversight that weakens the defensibility of the numerical comparisons. In the revised manuscript we have added a new paragraph in the Numerical results section that specifies the optimization method (a scalarized multi-objective evolutionary algorithm), population size, and total function-evaluation budget. To address the dimensionality bias, we have re-run all optimizations with the evaluation budget scaled linearly with the number of free parameters, thereby granting the full O(d) protocol substantially more search effort than the sparse ansatzes. The updated Pareto fronts and hypervolume values are now presented; they confirm that the sparse ansatzes retain their advantage, especially when the objective is to minimize phase count. These controls directly respond to the referee's concern and reinforce the central representational claim. revision: yes

Circularity Check

0 steps flagged

No circularity: results from direct numerical optimization of phase subsets

full rationale

The paper's central claims rest on numerical optimization of sparse SNAP phase subsets for fidelity objectives in bosonic qudit systems, with comparisons via hypervolume of Pareto fronts. No derivation chain reduces a claimed prediction or first-principles result to its own inputs by construction, nor does any load-bearing step rely on self-citation of an unverified uniqueness theorem or ansatz. The approach is self-contained against external benchmarks through explicit simulation of ideal and noisy regimes for target states up to d=64.

Axiom & Free-Parameter Ledger

1 free parameters · 0 axioms · 0 invented entities

The work relies on numerical optimization rather than new physical axioms or entities; free parameters are the selected phase subsets that are optimized for each target.

free parameters (1)
  • subset of SNAP phases
    The sparse ansatzes optimize only chosen subsets of phases instead of the full set; values are determined by the multi-objective optimizer for each target state.

pith-pipeline@v0.9.0 · 5817 in / 1126 out tokens · 57307 ms · 2026-05-21T10:54:17.213826+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

45 extracted references · 45 canonical work pages

  1. [1]

    Lloyd and S

    S. Lloyd and S. L. Braunstein, Quantum computation over continuous variables, Phys. Rev. Lett.82, 1784 (1999)

    work page 1999

  2. [2]

    R. W. Heeres, B. Vlastakis, E. Holland, S. Krastanov, V. V. Albert, L. Frunzio, L. Jiang, and R. J. Schoelkopf, Cavity state manipulation using photon-number selective phase gates, Phys. Rev. Lett.115, 137002 (2015)

    work page 2015

  3. [3]

    Krastanov, V

    S. Krastanov, V. V. Albert, C. Shen, C.-L. Zou, R. W. Heeres, B. Vlastakis, R. J. Schoelkopf, and L. Jiang, Uni- versal control of an oscillator with dispersive coupling to a qubit, Phys. Rev. A92, 040303 (2015)

    work page 2015

  4. [4]

    Chakram, K

    S. Chakram, K. He, A. V. Dixit, A. E. Oriani, R. K. Naik, N. Leung, H. Kwon, W.-L. Ma, L. Jiang, and D. I. Schus- ter, Multimode photon blockade, Nature Physics18, 879 (2022)

    work page 2022

  5. [5]

    Krayzman, C

    L. Krayzman, C. U. Lei, S. Ganjam, J. Teoh, L. Frunzio, and R. J. Schoelkopf, Superconducting quantum mem- ory with a suspended coaxial resonator, Applied Physics Letters124, 10.1063/5.0203906 (2024)

    work page doi:10.1063/5.0203906 2024

  6. [6]

    A. E. Oriani, F. Zhao, T. Roy, A. Anferov, K. He, A. Agrawal, R. Banerjee, S. Chakram, and D. I. Schus- ter, Niobium coaxial cavities with internal quality factors exceeding 1.4×10 9 for circuit quantum electrodynamics, Phys. Rev. Appl.24, 044080 (2025)

    work page 2025

  7. [7]

    T. Roy, T. Kim, A. Romanenko, and A. Grassellino, Qudit-based quantum computing with srf cavities at fer- milab (Fermi National Accelerator Laboratory (FNAL), Batavia, IL (United States), 2024)

    work page 2024

  8. [8]

    Bornman, T

    N. Bornman, T. Roy, J. A. Job, N. Anand, G. N. Perdue, S. Zorzetti, and M. S. Alam, Benchmarking the perfor- mance of a high-q cavity qudit using random unitaries, Quantum Science and Technology10, 025062 (2025)

    work page 2025

  9. [9]

    T. Kim, T. Roy, X. You, A. C. Li, H. Lamm, O. Pronitchev, M. Bal, S. Garattoni, F. Crisa, D. Bafia, et al., Ultracoherent superconducting cavity-based multi- qudit platform with error-resilient control, arXiv preprint arXiv:2506.03286 (2025)

    work page arXiv 2025

  10. [10]

    Prakash and T

    S. Prakash and T. Saha, Low Overhead Qutrit Magic State Distillation, Quantum9, 1768 (2025)

    work page 2025

  11. [11]

    Wills, M.-H

    A. Wills, M.-H. Hsieh, and H. Yamasaki, Constant- overhead magic state distillation, Nature Physics21, 1842 (2025)

    work page 2025

  12. [12]

    E. O. Kiktenko, A. S. Nikolaeva, and A. K. Fedorov, Colloquium: Qudits for decomposing multiqubit gates and realizing quantum algorithms, Rev. Mod. Phys.97, 021003 (2025)

    work page 2025

  13. [13]

    Y. Wang, Z. Hu, B. C. Sanders, and S. Kais, Qudits and high-dimensional quantum computing, Frontiers in PhysicsVolume 8 - 2020, 10.3389/fphy.2020.589504 (2020)

    work page doi:10.3389/fphy.2020.589504 2020

  14. [14]

    Jankovi´ c, J.-G

    D. Jankovi´ c, J.-G. Hartmann, M. Ruben, and P.-A. Hervieux, Noisy qudit vs multiple qubits: conditions on gate efficiency for enhancing fidelity, npj Quantum Infor- mation10, 59 (2024)

    work page 2024

  15. [15]

    Ogunkoya, K

    O. Ogunkoya, K. Morris, and D. M. K¨ urk¸ c¨ uo˜ glu, Inves- tigating parameter trainability in the snap-displacement protocol of a qudit system, Physica Scripta100, 075109 (2025)

    work page 2025

  16. [16]

    Royer, S

    B. Royer, S. Singh, and S. Girvin, Encoding qubits in multimode grid states, PRX Quantum3, 010335 (2022)

    work page 2022

  17. [17]

    Y. Xu, Y. Wang, E.-J. Kuo, and V. V. Albert, Qubit- oscillator concatenated codes: Decoding formalism and code comparison, PRX Quantum4, 020342 (2023)

    work page 2023

  18. [18]

    B. L. Brock, S. Singh, A. Eickbusch, V. V. Sivak, A. Z. Ding, L. Frunzio, S. M. Girvin, and M. H. Devoret, Quan- tum error correction of qudits beyond break-even, Nature 641, 612 (2025)

    work page 2025

  19. [19]

    D. M. K¨ urk¸ c¨ uoglu, H. Lamm, O. Ogunkoya, and L. Pier- attelli, The 2t-quoctit: a two-mode bosonic qudit for high energy physics, arXiv preprint arXiv:2509.21941 (2025)

    work page arXiv 2025

  20. [20]

    X. Deng, S. Li, Z.-J. Chen, Z. Ni, Y. Cai, J. Mai, L. Zhang, P. Zheng, H. Yu, C.-L. Zou,et al., Quantum- enhanced metrology with large fock states, Nature Physics20, 1874 (2024)

    work page 2024

  21. [21]

    F. Zhao, Z. Li, A. V. Dixit, T. Roy, A. Vrajitoarea, R. Banerjee, A. Anferov, K.-H. Lee, D. I. Schuster, and A. Chou, Flux-tunable cavity for dark matter detection, Phys. Rev. Lett.135, 201002 (2025)

    work page 2025

  22. [22]

    Agrawal, A

    A. Agrawal, A. V. Dixit, T. Roy, S. Chakram, K. He, R. K. Naik, D. I. Schuster, and A. Chou, Stimulated emission of signal photons from dark matter waves, Phys. Rev. Lett.132, 140801 (2024)

    work page 2024

  23. [23]

    R. W. Heeres, P. Reinhold, N. Ofek, L. Frunzio, L. Jiang, M. H. Devoret, and R. J. Schoelkopf, Implementing a uni- versal gate set on a logical qubit encoded in an oscillator, Nature Communications8, 94 (2017)

    work page 2017

  24. [24]

    Eickbusch, V

    A. Eickbusch, V. Sivak, A. Z. Ding, S. S. Elder, S. R. Jha, J. Venkatraman, B. Royer, S. M. Girvin, R. J. Schoelkopf, and M. H. Devoret, Fast universal control of an oscillator with weak dispersive coupling to a qubit, Nature Physics 18, 1464 (2022)

    work page 2022

  25. [25]

    X. You, A. C. Li, T. Roy, S. Zhu, A. Romanenko, A. Grassellino, Y. Lu, and S. Chakram, Floquet- engineered fast snap gates in weakly coupled circuit-qed systems, Phys. Rev. Appl.24, 034072 (2025)

    work page 2025

  26. [26]

    Kudra, M

    M. Kudra, M. Kervinen, I. Strandberg, S. Ahmed, M. Scigliuzzo, A. Osman, D. P. Lozano, M. O. Thol´ en, R. Borgani, D. B. Haviland, G. Ferrini, J. Bylander, A. F. Kockum, F. Quijandr´ ıa, P. Delsing, and S. Gasparinetti, Robust preparation of wigner-negative states with opti- mized snap-displacement sequences, PRX Quantum3, 030301 (2022)

    work page 2022

  27. [27]

    F¨ osel, S

    T. F¨ osel, S. Krastanov, F. Marquardt, and L. Jiang, Efficient cavity control with snap gates, arXiv preprint arXiv:2004.14256 (2020)

    work page arXiv 2004

  28. [28]

    Job, Efficient, direct compilation of su (n) opera- tions into snap & displacement gates, arXiv preprint arXiv:2307.11900 (2023)

    J. Job, Efficient, direct compilation of su (n) opera- tions into snap & displacement gates, arXiv preprint arXiv:2307.11900 (2023)

    work page arXiv 2023

  29. [29]

    Champion, Z

    E. Champion, Z. Wang, R. W. Parker, and M. S. Blok, Efficient control of a transmon qudit using effective spin- 7/2 rotations, Phys. Rev. X15, 021096 (2025)

    work page 2025

  30. [30]

    D. M. K¨ urk¸ c¨ uoglu, H. Lamm, and A. Maestri, Qudit gate decomposition dependence for lattice gauge theo- ries, arXiv preprint arXiv:2410.16414 (2024)

    work page arXiv 2024

  31. [31]

    Landgraf, C

    J. Landgraf, C. Fl¨ uhmann, T. F¨ osel, F. Marquardt, and R. J. Schoelkopf, Fast quantum control of cavities using an improved protocol without coherent errors, Phys. Rev. Lett.133, 260802 (2024)

    work page 2024

  32. [32]

    Pietik¨ ainen, O

    I. Pietik¨ ainen, O. c. v.ˇCernot´ ık, A. Eickbusch, A. Maiti, J. W. Garmon, R. Filip, and S. M. Girvin, Strategies and trade-offs for controllability and memory time of ultra- high-quality microwave cavities in circuit quantum elec- 14 trodynamics, PRX Quantum5, 040307 (2024)

    work page 2024

  33. [33]

    M. J. Cervia, H. Lamm, D. Liu, E. M. Murairi, and S. Zhu, Magic state distillation using asymptotically good codes on qudits, arXiv preprint arXiv:2512.21874 (2025). [34]https://github.com/qcpolimi/SparseSDAnzatz.git

    work page arXiv 2025

  34. [34]

    T. Roy, T. Kim, A. Romanenko, and A. Grassellino, Qudit-based quantum computing with SRF cavities at Fermilab, PoSLATTICE2023, 127 (2024)

    work page 2024

  35. [35]

    R. J. Glauber, Coherent and incoherent states of the ra- diation field, Phys. Rev.131, 2766 (1963)

    work page 1963

  36. [36]

    Bazgan, S

    C. Bazgan, S. Ruzika, C. Thielen, and D. Vanderpooten, The power of the weighted sum scalarization for approxi- mating multiobjective optimization problems, Theory of Computing Systems66, 395 (2022)

    work page 2022

  37. [37]

    Bergstra, R

    J. Bergstra, R. Bardenet, Y. Bengio, and B. K´ egl, Al- gorithms for hyper-parameter optimization, inAdvances in Neural Information Processing Systems (NeurIPS) (2011) pp. 2546–2554

    work page 2011

  38. [38]

    A. P. Guerreiro, C. M. Fonseca, and L. Paquete, The hypervolume indicator: Computational problems and al- gorithms, ACM Comput. Surv.54, 119:1 (2022)

    work page 2022

  39. [39]

    Auger, J

    A. Auger, J. Bader, D. Brockhoff, and E. Zitzler, Hypervolume-based multiobjective optimization: The- oretical foundations and practical implications, Theor. Comput. Sci.425, 75 (2012)

    work page 2012

  40. [40]

    Miettinen,Nonlinear Multiobjective Optimization, In- ternational Series in Operations Research & Management Science (Springer, 1999)

    K. Miettinen,Nonlinear Multiobjective Optimization, In- ternational Series in Operations Research & Management Science (Springer, 1999)

    work page 1999

  41. [41]

    Mezzadri, How to generate random matrices from the classical compact groups, arXiv preprint math- ph/0609050 (2006)

    F. Mezzadri, How to generate random matrices from the classical compact groups, arXiv preprint math- ph/0609050 (2006). 15 Appendix A: Fock and superposition state construction Here we list the optimized sequence for preparing var- ious useful Fock and superposition states starting from the ground state using single-parameter SNAP gates in Table III. The ...

    work page arXiv 2006

  42. [42]

    Pruning Mask To further reduce the number of learnable phases, we introduce an optional masking rule that removes all learnable phases below a tunable threshold line. The pur- pose of this rule is to avoid placing learnable phases in regions that are unlikely to influence the quantum state, such as high Fock states during early blocks because it is likely...

    work page 2005

  43. [43]

    The channel is represented using a Kraus decomposition of the form: E(ρ) = ℓmaxX ℓ=0 KℓρK † ℓ (H1) whereK ℓ are the Kraus operators describing the loss of ℓphotons

    Photon-Loss Model Photon-loss is modeled as a trace-preserving quantum channel acting after each block of the SNAP displace- ment protocol. The channel is represented using a Kraus decomposition of the form: E(ρ) = ℓmaxX ℓ=0 KℓρK † ℓ (H1) whereK ℓ are the Kraus operators describing the loss of ℓphotons. Each Kraus operator is given by: Kℓ = r (1−η) ℓ ℓ! ·...

    work page

  44. [44]

    The duration of a displacement gate is very short compared to the cavity lifetime and can be safely assumed to beTD ≃0.1µs [27]

    Gate Duration Estimation The simulation of photon-loss noise requires to esti- mate the duration of each block,T block. The duration of a displacement gate is very short compared to the cavity lifetime and can be safely assumed to beTD ≃0.1µs [27]. The duration of a SNAP gate instead depends strongly on the number of phases that need to be applied because...

    work page

  45. [45]

    Results In this section we report the results obtained when learning the gate parameters under noisy conditions. Due to the very large computational cost of running the nu- merical simulations, in particular the hyperparameter op- timization phase, we choose to rely on the optimal hy- perparameters identified in noiseless conditions. Note, however, that a...

    work page