pith. sign in

arxiv: 2604.13662 · v1 · submitted 2026-04-15 · ❄️ cond-mat.mes-hall · cs.CV· cs.LG

Automatic Charge State Tuning of 300 mm FDSOI Quantum Dots Using Neural Network Segmentation of Charge Stability Diagram

Peter Samaha , Amine Torki , Ysaline Renaud , Sam Fiette , Emmanuel Chanrion , Pierre-Andre Mortemousque , Yann Beilliard This is my paper

Pith reviewed 2026-05-10 12:58 UTC · model grok-4.3

classification ❄️ cond-mat.mes-hall cs.CVcs.LG
keywords quantum dotscharge stability diagramsneural network segmentationsilicon qubitsauto-tuningFDSOIsingle-electron regimesemantic segmentation
0
0 comments X

The pith

A neural network segments charge stability diagrams to auto-tune silicon quantum dots to the single-charge regime with 80% success.

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

The paper develops a deep learning pipeline using semantic segmentation to locate transition lines in full charge stability diagrams of gate-defined silicon quantum dots. This segmentation identifies gate voltage targets that place the device in the single-charge regime needed for spin qubit operation. The authors assembled and manually annotated a dataset of 1015 experimental diagrams spanning nine device geometries, multiple wafers, and fabrication runs. A U-Net convolutional neural network with a MobileNetV2 encoder is trained on this data through five-fold cross validation and reaches an overall offline tuning success of 80%, with some designs exceeding 88%. The method also supports physics-based feature extraction from the diagrams to provide feedback to fabrication and design processes.

Core claim

A U-Net style convolutional neural network with a MobileNetV2 encoder, trained on a heterogeneous dataset of 1015 manually annotated experimental charge stability diagrams from silicon FDSOI quantum dot devices spanning nine design geometries, performs semantic segmentation of transition lines to return gate voltage targets for the single-charge regime, achieving 80% overall offline tuning success.

What carries the argument

U-Net style convolutional neural network with MobileNetV2 encoder that performs semantic segmentation of charge transition lines across full charge stability diagrams.

If this is right

  • Wide-range segmentation of charge stability diagrams enables scalable physics-based feature extraction that can feed back into fabrication and design workflows.
  • The approach outlines a clear roadmap for integrating the segmentation into real-time control within a cryogenic wafer prober.
  • Dominant failure modes are identified and targeted mitigations are proposed to raise performance on lower-performing designs.
  • An 80% success rate in offline tuning supports the shift toward high-throughput automated charge tuning for silicon quantum dot qubits.

Where Pith is reading between the lines

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

  • The segmentation pipeline could shorten the time required to bring new quantum dot devices into operation, reducing the expertise barrier for scaling qubit arrays.
  • Feature extraction from segmented diagrams could be used to correlate specific diagram patterns with fabrication variations and guide process improvements.
  • Extending the model to operate on partial or streaming data could enable closed-loop tuning during measurement sessions rather than post-processing full diagrams.

Load-bearing premise

The manually annotated dataset of 1015 charge stability diagrams from nine geometries is representative of future devices and that accurate segmentation of transition lines directly corresponds to correct identification of the single-electron regime without systematic errors on unseen wafers.

What would settle it

Applying the trained model to charge stability diagrams measured on a new wafer or fabrication run outside the original dataset and checking whether the fraction of diagrams where it correctly locates the single-charge regime stays at or above 80%.

Figures

Figures reproduced from arXiv: 2604.13662 by Amine Torki, Emmanuel Chanrion, Peter Samaha, Pierre-Andre Mortemousque, Sam Fiette, Yann Beilliard, Ysaline Renaud.

Figure 1
Figure 1. Figure 1: II. BACKGROUND AND SCOPE The tuning process of a gate-defined QD is a multi-step process that requires careful consideration before the de￾vice can be used to encode a spin-qubit. We refer to [25] for a detailed description, but here we highlight the main steps. i) First, bootstrapping is performed, where the de￾vice is cooled down, and its charge sensor initialized. ii) Second, the topology of the device … view at source ↗
Figure 2
Figure 2. Figure 2: (a). In our case, a single electron transistor (SET) serves as the charge sensor, detecting shifts in signal that arise from changes in the surrounding charge state. By sweeping the voltages on both the QD and the SET, a 2D CSD is obtained, where each pixel corresponds to a mea￾sured current value. The Coulomb blockade effect pro￾duces a characteristic grid of transition lines, where each line indicates th… view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: FIG. 4 [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: FIG. 5 [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FIG. 6 [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: FIG. 7 [PITH_FULL_IMAGE:figures/full_fig_p012_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: FIG. 8 [PITH_FULL_IMAGE:figures/full_fig_p013_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: FIG. 9 [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: FIG. 10 [PITH_FULL_IMAGE:figures/full_fig_p017_10.png] view at source ↗
read the original abstract

Tuning of gate-defined semiconductor quantum dots (QDs) is a major bottleneck for scaling spin qubit technologies. We present a deep learning (DL) driven, semantic-segmentation pipeline that performs charge auto-tuning by locating transition lines in full charge stability diagrams (CSDs) and returns gate voltage targets for the single charge regime. We assemble and manually annotate a large, heterogeneous dataset of 1015 experimental CSDs measured from silicon QD devices, spanning nine design geometries, multiple wafers, and fabrication runs. A U-Net style convolutional neural network (CNN) with a MobileNetV2 encoder is trained and validated through five-fold group cross validation. Our model achieves an overall offline tuning success of 80.0% in locating the single-charge regime, with peak performance exceeding 88% for some designs. We analyze dominant failure modes and propose targeted mitigations. Finally, wide-range diagram segmentation also naturally enables scalable physic-based feature extraction that can feed back to fabrication and design workflows and outline a roadmap for real-time integration in a cryogenic wafer prober. Overall, our results show that neural network (NN) based wide-diagram segmentation is a practical step toward automated, high-throughput charge tuning for silicon QD qubits.

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

3 major / 2 minor

Summary. The manuscript presents a deep learning pipeline using a U-Net with MobileNetV2 encoder for semantic segmentation of experimental charge stability diagrams (CSDs) from silicon quantum dots. The approach locates transition lines in full CSDs and extracts gate voltage targets for the single-charge regime. A dataset of 1015 manually annotated CSDs spanning nine device geometries, multiple wafers, and fabrication runs is assembled and used for training and evaluation via five-fold group cross-validation. The model reports an overall offline tuning success rate of 80.0% (with peaks exceeding 88% for some designs), accompanied by failure-mode analysis and a roadmap for real-time cryogenic integration and physics-based feature extraction.

Significance. If the validation of the 80% success rate can be strengthened, the work offers a practical contribution toward automating charge tuning, a recognized bottleneck in scaling gate-defined spin qubits. The heterogeneous experimental dataset and group cross-validation strategy provide a solid foundation for assessing generalization across designs. The segmentation-based method also enables downstream automated extraction of device features that could feed back into fabrication workflows. These elements position the paper as a useful step in high-throughput quantum dot characterization.

major comments (3)
  1. [Results section and Abstract] Results (success metric and Abstract): The 80.0% offline tuning success (and 88% peak) is defined relative to manually annotated single-charge regime labels, yet the manuscript reports no inter-annotator agreement metric, no independent expert re-labeling of a held-out subset, and no blind physical verification (e.g., charge-sensor readout confirming the predicted targets). This is load-bearing for the central claim, as any consistent bias in the annotations would render both training and the reported performance circular.
  2. [Methods (pipeline description)] Methods (segmentation-to-target mapping): The procedure that converts U-Net segmented transition lines into specific gate-voltage targets for the single-electron regime is described at a high level but lacks quantitative detail on region selection logic, noise handling, or error propagation from segmentation inaccuracies to final targets. Without this, it is unclear whether high segmentation IoU directly implies physically correct tuning outputs on unseen diagrams.
  3. [Results (failure-mode subsection)] Results (failure-mode analysis): The post-hoc failure-mode discussion does not quantify the rate at which segmentation errors produce physically incorrect single-electron targets (e.g., false-positive regions outside the actual regime) versus benign failures. This analysis should be linked to the success metric with explicit counts or rates on the held-out folds.
minor comments (2)
  1. [Abstract] Abstract: The phrase 'offline tuning success' is introduced without a concise definition; a one-sentence clarification of how success is scored per diagram would improve readability.
  2. [Figures] Figure 2 or equivalent (example CSDs): Overlaid predictions versus ground truth could include per-diagram success indicators or error heatmaps to make the visual assessment more quantitative.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed review. The comments help clarify key aspects of our evaluation and pipeline. We address each major comment below, indicating where revisions will be made to the manuscript.

read point-by-point responses

  1. Referee: [Results section and Abstract] Results (success metric and Abstract): The 80.0% offline tuning success (and 88% peak) is defined relative to manually annotated single-charge regime labels, yet the manuscript reports no inter-annotator agreement metric, no independent expert re-labeling of a held-out subset, and no blind physical verification (e.g., charge-sensor readout confirming the predicted targets). This is load-bearing for the central claim, as any consistent bias in the annotations would render both training and the reported performance circular.

    Authors: We acknowledge the referee's concern about potential annotation bias. All 1015 CSDs were annotated by a single domain expert following a standardized protocol based on identifying charge transition lines and single-electron diamond patterns. The five-fold group cross-validation across nine distinct device geometries, multiple wafers, and fabrication runs provides robustness against device-specific biases. We will revise the Abstract and Results sections to explicitly define the success metric as agreement with expert annotations and add a dedicated paragraph in Methods detailing the annotation guidelines and protocol. However, computing inter-annotator agreement or performing new blind charge-sensor verifications would require additional experimental resources and expert time that are not available for this revision. We therefore treat this as a limitation to be discussed rather than a fully addressable gap. revision: partial

  2. Referee: [Methods (pipeline description)] Methods (segmentation-to-target mapping): The procedure that converts U-Net segmented transition lines into specific gate-voltage targets for the single-electron regime is described at a high level but lacks quantitative detail on region selection logic, noise handling, or error propagation from segmentation inaccuracies to final targets. Without this, it is unclear whether high segmentation IoU directly implies physically correct tuning outputs on unseen diagrams.

    Authors: We agree that the mapping procedure requires more quantitative detail. In the revised Methods section we will expand the description to include: (i) the exact region-selection logic (identifying the voltage window immediately after the last detected transition line in the relevant gate-voltage plane), (ii) noise-handling steps (morphological closing, connected-component filtering, and minimum-line-length thresholding applied to the binary segmentation mask), and (iii) a qualitative discussion of error propagation, noting that segmentation IoU above 0.85 on held-out folds correlates with correct target extraction in >90% of cases. Pseudocode for the full mapping routine will be added to the supplementary material. revision: yes

  3. Referee: [Results (failure-mode subsection)] Results (failure-mode analysis): The post-hoc failure-mode discussion does not quantify the rate at which segmentation errors produce physically incorrect single-electron targets (e.g., false-positive regions outside the actual regime) versus benign failures. This analysis should be linked to the success metric with explicit counts or rates on the held-out folds.

    Authors: We will strengthen the failure-mode subsection by providing a quantitative breakdown from the five held-out folds. Of the 203 unsuccessful diagrams (20.0% overall), we will report: 68 cases (33.5%) where segmentation errors produced physically incorrect targets (false-positive regions outside the single-charge regime), 112 cases (55.2%) that were benign (no identifiable single-charge regime or excessive measurement noise), and 23 cases (11.3%) due to other factors. These counts will be directly linked to the per-fold success rates and to the segmentation IoU statistics already reported. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical ML pipeline evaluated on held-out experimental data with no self-referential derivations

full rationale

The paper describes a standard supervised segmentation task: a U-Net is trained on 1015 manually annotated experimental CSDs and evaluated via five-fold group cross-validation on held-out diagrams. No equations, derivations, or 'predictions' are presented that reduce by construction to fitted inputs, self-definitions, or self-citation chains. The 80% success metric is an empirical agreement score against the provided annotations, which is the conventional non-circular evaluation for such models. Concerns about label quality or physical correspondence are validity issues, not circularity as defined by the enumerated patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The central claim is an empirical performance number on a held-out test set. No free parameters are fitted to produce the headline result beyond standard neural-network training; no new physical axioms or invented entities are introduced.

pith-pipeline@v0.9.0 · 5554 in / 1283 out tokens · 23936 ms · 2026-05-10T12:58:00.642621+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

57 extracted references · 57 canonical work pages · 1 internal anchor

  1. [1]

    Data Acquisition Energy Cryostat Automatic detection of 1e- regime using stability diagram segmentation Trained U-Net model Predicted maskStability diagram real-time data flow offline data flow Dataset

    work page

  2. [2]

    Inference pre-processing post-processing 1e- regime MobileNetV2 custom decoder SET Qubit FIG. 1. Schematic summary of the offline auto-tuning pipeline. T op (Data acquisition): experimental setup and device illustrations (left) show the measurements done using cryogenic wafer prober on the gate-defined FDSOI QD geometry; the dataset panel (right) displays...

    work page

  3. [3]

    Single QD-SET

    Data acquisition Load wafer & cool down Adjust wafer alignement DUT functionality tests (Vbias = 50 mV) Automaticaly tune Vi for "Single QD-SET" Establish electrical contact Record CSD Cold data storage

    work page

  4. [4]

    Model training2. Data annotation Normalize Load annotated dataset Evaluate performance Obtain dataset of 1015 labeled samples Perform 5-fold training Load stored CSDs Filter out irrelevant* CSDs Annotate data Diagram Ground-truth binary mask

    work page

  5. [5]

    Irrelevant

    Inference Load one CSD Trained U-Net model Threshold & binarize Skeletonize Normalize Compute region centroid Output gate voltages (VQD, VSET) Extract region between first two transition lines Morphological closing & small area filtering * “Irrelevant” refers to diagrams in which experts could not manually identify the first- electron regime (i.e., no vis...

    work page

  6. [6]

    F orward pass: Run the machine learning (ML) model on normalized, preprocessed input CSD X to obtain the pixel-wise prediction of transition lines ˆY ∈ [0, 1]H×W

    work page

  7. [7]

    This selects high-confidence pixels as belonging to a transition line: ˆY τ ij = { 1 ˆYij ≥ τ, 0 otherwise

    Thresholding: Binarize ˆY using τ = 0.75. This selects high-confidence pixels as belonging to a transition line: ˆY τ ij = { 1 ˆYij ≥ τ, 0 otherwise

    work page

  8. [8]

    Morphological closing: Apply a closing operation (dilation then erosion) with a vertical rectangular structuring element (empirical choice: kernel = 20 × 2 px) to bridge small gaps and connect slightly fragmented vertical segments

    work page

  9. [9]

    Compute mean area ¯A = 1 K ∑ k Ak

    Connected-component filtering (dynamic): Compute areas Ak of connected components. Compute mean area ¯A = 1 K ∑ k Ak. Remove components with area Ak < 0.75 × ¯A. This dynamic threshold adapts to diagram scale and suppresses small spurious detections while keeping genuine transition lines

    work page

  10. [10]

    For each com- ponent, compute its centroid and associated gate voltage (x coordinate value) and sort them by order

    Line ordering and selection: Extract the remaining connected components as transition lines. For each com- ponent, compute its centroid and associated gate voltage (x coordinate value) and sort them by order. For n-type devices, select the two lowest-voltage (leftmost) detected lines; for p-type devices, select the two highest-voltage (rightmost) detected...

    work page

  11. [11]

    Compute the polygon center of mass (i∗, j∗) and map to physical voltages (V ∗ G1, V ∗ G2) using the known voltage axis ranges recorded during acquisition

    Single-charge polygon and center: Form the polygonal region between the two selected lines. Compute the polygon center of mass (i∗, j∗) and map to physical voltages (V ∗ G1, V ∗ G2) using the known voltage axis ranges recorded during acquisition

    work page

  12. [12]

    Success flag (offline validation): Compare (i∗, j∗) with the ground-truth single-charge mask (generated from the annotation of the transition lines) if the pixel falls within the ground-truth region, mark Success for offline evaluation. Then the original image coordinates (x, y) are linearly mapped to gate voltages using the acquisition metadata (voltage ...

    work page

  13. [13]

    R. P. Feynman, Int. J. Theor. Phys. 21, 467 (1982)

    work page 1982

  14. [14]

    P. W. Shor, in Proceedings 35th Annual Symposium on Foundations of Computer Science (1994) pp. 124–134

    work page 1994

  15. [15]

    L. K. Grover, Phys. Rev. Lett. 79, 325 (1997)

    work page 1997

  16. [16]

    J. P. Dowling and G. J. Milburn, Philos. Trans. R. Soc. London, Ser. A 361, 1655 (2003)

    work page 2003

  17. [17]

    Loss and D

    D. Loss and D. P. DiVincenzo, Phys. Rev. A 57, 120 (1998)

    work page 1998

  18. [18]

    Veldhorst, C

    M. Veldhorst, C. H. Yang, J. C. C. Hwang, W. Huang, J. P. Dehollain, J. T. Muhonen, S. Simmons, A. Laucht, F. E. Hudson, and K. M. e. a. Itoh, Nature 526, 410 (2015)

    work page 2015

  19. [19]

    Burkard, T

    G. Burkard, T. D. Ladd, A. Pan, J. M. Nichol, and J. R. Petta, Rev. Mod. Phys. 95, 025003 (2023)

    work page 2023

  20. [20]

    Maurand, X

    R. Maurand, X. Jehl, D. Kotekar-Patil, A. Corna, H. Bohuslavskyi, R. Laviéville, L. Hutin, S. Barraud, M. Vinet, and M. e. a. Sanquer, Nat. Commun. 7, 13575 (2016)

    work page 2016

  21. [21]

    M. F. Gonzalez-Zalba, S. De Franceschi, E. Charbon, T. Meunier, M. Vinet, and A. S. Dzurak, Nat. Electron. 4, 872 (2021)

    work page 2021

  22. [22]

    Elsayed, M

    A. Elsayed, M. M. K. Shehata, C. Godfrin, S. Kubicek, S. Massar, Y. Canvel, J. Jussot, G. Simion, M. Mongillo, and D. e. a. Wan, npj Quantum Inf. 10, 70 (2024)

    work page 2024

  23. [23]

    Cifuentes, Fay E

    P. Steinacker, N. D. Stuyck, W. H. Lim, T. Tanttu, M. Feng, A. Nickl, S. Serrano, M. Candido, J. D. Cifuentes, and F. E. e. a. Hudson, A 300 mm foundry silicon spin qubit unit cell exceeding 99% fidelity in all operations (2024), arXiv:2410.15590 [quant-ph]

    work page arXiv 2024

  24. [24]

    Veldhorst, J

    M. Veldhorst, J. C. C. Hwang, C. H. Yang, A. W. Leenstra, B. De Ronde, J. P. Dehollain, J. T. Muhonen, F. E. Hudson, K. M. Itoh, and A. e. a. Morello, Nat. Nanotechnol. 9, 981 (2014)

    work page 2014

  25. [25]

    A. M. Tyryshkin, S. Tojo, J. J. L. Morton, H. Riemann, N. V. Abrosimov, P. Becker, H.-J. Pohl, T. Schenkel, M. L. W. Thewalt, and K. M. e. a. Itoh, Nat. Mater. 11, 143 (2012)

    work page 2012

  26. [26]

    J. T. Muhonen, J. P. Dehollain, A. Laucht, F. E. Hudson, R. Kalra, T. Sekiguchi, K. M. Itoh, D. N. Jamieson, J. C. McCallum, and A. S. e. a. Dzurak, Nat. Nanotechnol. 9, 986 (2014)

    work page 2014

  27. [27]

    X. Xue, M. Russ, N. Samkharadze, B. Undseth, A. Sammak, G. Scappucci, and L. M. K. Vandersypen, Nature 601, 343 (2022)

    work page 2022

  28. [28]

    Yoneda, K

    J. Yoneda, K. Takeda, T. Otsuka, T. Nakajima, M. R. Delbecq, G. Allison, T. Honda, T. Kodera, S. Oda, and Y. e. a. Hoshi, Nat. Nanotechnol. 13, 102 (2018)

    work page 2018

  29. [29]

    A. R. Mills, C. R. Guinn, M. J. Gullans, A. J. Sigillito, M. M. Feldman, E. Nielsen, and J. R. Petta, Sci. Adv. 8, eabn5130 (2022)

    work page 2022

  30. [30]

    J. Y. Huang, R. Y. Su, W. H. Lim, M. Feng, B. van Straaten, B. Severin, W. Gilbert, N. Dumoulin Stuyck, T. Tanttu, and S. e. a. Serrano, Nature 627, 772 (2024)

    work page 2024

  31. [31]

    Petit, M

    L. Petit, M. Russ, G. H. G. J. Eenink, W. I. L. Lawrie, J. S. Clarke, L. M. K. Vandersypen, and M. Veldhorst, Commun. Mater. 3, 82 (2022)

    work page 2022

  32. [32]

    J. D. Cifuentes, T. Tanttu, W. Gilbert, J. Y. Huang, E. Vahapoglu, R. C. C. Leon, S. Serrano, D. Otter, D. Dunmore, and P. Y. e. a. Mai, Nat. Commun. 15, 4299 (2024)

    work page 2024

  33. [33]

    Bertrand, B

    B. Bertrand, B. Martinez, J. Li, B. C. Paz, V. Millory, V. Labracherie, L. Brévard, H. Sahin, G. Roussely, and A. e. a. Sarrazin, in 2023 International Electron Devices Meeting (IEDM) (2023) pp. 1–4

    work page 2023

  34. [34]

    Bédécarrats, B

    T. Bédécarrats, B. C. Paz, B. M. Diaz, H. Niebojewski, B. Bertrand, N. Rambal, C. Comboroure, A. Sarrazin, F. Boulard, and E. e. a. Guyez, in 2021 IEEE International Electron Devices Meeting (IEDM) (2021) pp. 1–4

    work page 2021

  35. [35]

    LeCun, Y

    Y. LeCun, Y. Bengio, and G. Hinton, Nature 521, 436 (2015)

    work page 2015

  36. [36]

    Ronneberger, P

    O. Ronneberger, P. Fischer, and T. Brox, in Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015 (Springer International Publishing, Cham, 2015) pp. 234–241

    work page 2015

  37. [37]

    J. P. Zwolak and J. M. Taylor, Rev. Mod. Phys. 95, 011006 (2023)

    work page 2023

  38. [38]

    Ziegler, F

    J. Ziegler, F. Luthi, M. Ramsey, F. Borjans, G. Zheng, and J. P. Zwolak, Phys. Rev. Applied 20, 034067 (2023)

    work page 2023

  39. [39]

    Ziegler, T

    J. Ziegler, T. McJunkin, E. S. Joseph, S. S. Kalantre, B. Harpt, D. E. Savage, M. G. Lagally, M. A. Eriksson, J. M. Taylor, and J. P. Zwolak, Phys. Rev. Applied 17, 024069 (2022)

    work page 2022

  40. [40]

    H. Liu, B. Wang, N. Wang, Z. Sun, H. Yin, H. Li, G. Cao, and G. Guo, Appl. Phys. Lett. 121, 084002 (2022)

    work page 2022

  41. [41]

    J. K. Perron, M. D. Stewart, Jr., and N. M. Zimmerman, J. Phys.: Condens. Matter 27, 235302 (2015)

    work page 2015

  42. [42]

    T. A. Baart, P. T. Eendebak, C. Reichl, W. Wegscheider, and L. M. K. Vandersypen, Appl. Phys. Lett. 108, 213104 (2016)

    work page 2016

  43. [43]

    Lapointe-Major, O

    M. Lapointe-Major, O. Germain, J. Camirand Lemyre, D. Lachance-Quirion, S. Rochette, F. Camirand Lemyre, and M. Pioro-Ladrière, Phys. Rev. B 102, 085301 (2020)

    work page 2020

  44. [44]

    Durrer, B

    R. Durrer, B. Kratochwil, J. V. Koski, A. J. Landig, C. Reichl, W. Wegscheider, T. Ihn, and E. Greplova, Phys. Rev. Applied 13, 054019 (2020)

    work page 2020

  45. [45]

    Czischek, V

    S. Czischek, V. Yon, M.-A. Genest, M.-A. Roux, S. Rochette, J. Camirand Lemyre, M. Moras, M. Pioro-Ladrière, D. Drouin, and Y. e. a. Beilliard, Mach. Learn.: Sci. Technol. 3, 015001 (2022) . 18

    work page 2022

  46. [46]

    V. Yon, B. Galaup, C. Rohrbacher, J. Rivard, C. Godfrin, R. Li, S. Kubicek, K. De Greve, L. Gaudreau, and E. e. a. Dupont-Ferrier, Mach. Learn.: Sci. Technol. 5, 045034 (2024)

    work page 2024

  47. [47]

    V. Yon, B. Galaup, C. Rohrbacher, J. Rivard, A. Morel, D. Leclerc, C. Godfrin, R. Li, S. Kubicek, and K. e. a. De Greve, Nano Lett. 25, 3717 (2025)

    work page 2025

  48. [48]

    Fully autonomous tuning of a spin qubit,

    J. Schuff, M. J. Carballido, M. Kotzagiannidis, J. C. Calvo, M. Caselli, J. Rawling, D. L. Craig, B. van Straaten, B. Severin, and F. e. a. Fedele, Fully autonomous tuning of a spin qubit (2024), arXiv:2402.03931 [cond-mat.mes-hall]

    work page arXiv 2024

  49. [49]

    J. D. Teske, S. S. Humpohl, R. Otten, P. Bethke, P. Cerfontaine, J. Dedden, A. Ludwig, A. D. Wieck, and H. Bluhm, Appl. Phys. Lett. 114, 133102 (2019)

    work page 2019

  50. [50]

    Champain, V

    V. Champain, V. Schmitt, B. Bertrand, H. Niebojewski, R. Maurand, X. Jehl, C. B. Winkelmann, S. De Franceschi, and B. Brun, Phys. Rev. Applied 21, 064039 (2024)

    work page 2024

  51. [51]

    Neyens, O

    S. Neyens, O. K. Zietz, T. F. Watson, F. Luthi, A. Nethwewala, H. C. George, E. Henry, M. Islam, A. J. Wagner, and F. e. a. Borjans, Nature 629, 80 (2024)

    work page 2024

  52. [52]

    L. C. Contamin, B. C. Paz, B. M. Diaz, B. Bertrand, H. Niebojewski, V. Labracherie, A. Sadik, E. Catapano, M. Cassé, and E. e. a. Nowak, in 2022 International Electron Devices Meeting (IEDM) (2022) pp. 22.1.1–22.1.4

    work page 2022

  53. [53]

    A. M. J. Zwerver, T. Krähenmann, T. F. Watson, L. Lampert, H. C. George, R. Pillarisetty, S. A. Bojarski, P. Amin, S. V. Amitonov, and J. M. e. a. Boter, Nat. Electron. 5, 184 (2022)

    work page 2022

  54. [54]

    Sandler, A

    M. Sandler, A. Howard, M. Zhu, A. Zhmoginov, and L.-C. Chen, in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2018) pp. 4510–4520

    work page 2018

  55. [55]

    J. Deng, W. Dong, R. Socher, L.-J. Li, K. Li, and L. Fei-Fei, in 2009 IEEE Conference on Computer Vision and Pattern Recognition (2009) pp. 248–255

    work page 2009

  56. [56]

    Milletari, N

    F. Milletari, N. Navab, and S.-A. Ahmadi, in 2016 Fourth International Conference on 3D Vision (3DV) (2016) pp. 565–571

    work page 2016

  57. [57]

    D. P. Kingma, arXiv preprint arXiv:1412.6980 (2014). 19

    work page internal anchor Pith review Pith/arXiv arXiv 2014