Pith. sign in

REVIEW 3 major objections 4 minor 49 references

Multitask Scanning Probe Microscopy

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper demonstrates a live, closed-loop multitask scanning probe microscope in which a multitask Gaussian process with a learned task-covariance matrix decides both the next measurement location and the next imaging protocol, so that a…

desk verdict Genuinely live closed-loop multimodal AFM, but the info-transfer claim leans on five paired seeds and a fitted-model statistic; worth refereeing after a held-out check. read the letter →

arxiv 2608.09104 v1 pith:BHK7OMMT submitted 2026-08-10 cond-mat.mtrl-sci cs.LGphysics.ins-det

Aditya Raghavan , Yu Liu , Ian Mercer , JP Maria , Sergei Kalinin This is my paper · ORCID
classification cond-mat.mtrl-scics.LGphysics.ins-det
keywords multitaskGaussianprocessintrinsiccoregionalizationmodelautonomousscanningprobemicroscopyactivelearningmeasurementmodalityselectioncomposition-spreadwaferDARTtappingmodeAFM
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

Scanning probe microscopy usually maps one property at a time, and multimodal studies either repeat every mode at every grid point or fix the protocol in advance. This paper claims that a microscope can instead treat each measurement mode as a related prediction task and, during the experiment, choose both where and how to measure next. The load-bearing idea is that a multitask Gaussian process can learn the spatial structure of each mode together with a cross-modal covariance, so a measurement in one mode updates the predicted map of the other mode across the entire wafer. The authors demonstrate this on a composition-spread AlScN wafer with tapping-mode and DART roughness as two tasks, using five paired seed measurements to initialize the relationship and 25 subsequent single-mode scans. If the claim holds, wafer-scale multimodal screening can be done with far fewer slow or damaging measurements.

What carries the argument

The central object is the intrinsic coregionalization model (ICM), a multitask Gaussian-process construction in which the covariance between two location-task pairs factorizes as $K[(x,t),(x',t')] = K_x(x,x')B_{tt'}$, with $B = WW^T + \mathrm{diag}(\kappa)$ the task-covariance matrix. $B$ is what transfers information: a measurement of task $t$ at position $x$ updates the posterior for task $t'$ at all positions through the learned cross-task entry $B_{tt'}$. The implementation pairs this with a random-scalarization upper-confidence-bound acquisition function that samples Dirichlet task weights and selects the location-task pair maximizing the weighted UCB score, plus a forced-exploration step every third iteration that picks the farthest unmeasured position with a random task. Paired seed measurements at the start provide the primary constraint on $B$.

What would settle it

Repeat the experiment with several different sets of paired seed positions (or with 10-20 paired seeds instead of five) and compare the learned $\rho$ and the final posterior maps against a dense ground-truth map of both modes; if $\rho$ swings widely across seed sets or the joint model predicts held-out co-located pairs no better than two independent Gaussian processes, the central transfer claim fails.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a multitask Gaussian process with an intrinsic coregionalization model can run live on an operating atomic force microscope and jointly reconstruct two response landscapes from non-coincident observations. The experiment acquired five paired seed measurements at randomly chosen positions, learned a task-covariance matrix whose normalized off-diagonal entry is $\rho = 0.752$, and then let a random-scalarization upper-confidence-bound acquisition function pick one location--task pair at a time. After 35 total scans (16 tapping, 19 DART), the joint model reduced the grid-averaged posterior standard deviation from roughly 0.925 nm to 0.176 nm for tapping and from 1.098 nm to 0.235 nm for DART, while only one modality was measured at each non-seed location. The paper is careful to note that this uncertainty reduction is a property of the fitted model and that the high posterior correlation between the two maps ($r=0.984$) is partly imposed by the shared spatial kernel.

Load-bearing premise

The whole information-transfer argument rests on the cross-task correlation being learned reliably from just five paired seed measurements; if those five pairs are unrepresentative or dominated by noise, the transferred predictions and the reported uncertainty reduction would not hold.

Editorial extensions

If this is right

  • A single measurement at each location can produce posterior maps for two protocols, eliminating the need to acquire both modes at every grid point.
  • Rapid, weakly perturbative modes can guide the selective deployment of slower contact or spectroscopic measurements, reducing tip wear and sample modification.
  • Every single-mode measurement improves the predicted landscape of the unmeasured mode, so knowledge accumulates across the whole wafer even when that mode is never run at most positions.
  • With the cost-aware acquisition function of Eq. 13, the controller can explicitly trade expected information gain against acquisition time, damage, and mode-switching overhead.

Reading between the lines

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

  • Editorial inference: for modality pairs that are only weakly correlated, five paired seed measurements may be too few to pin down $B$; a practical rule would be to acquire more paired seeds when the expected cross-task correlation is low or the per-task noise is high.
  • Editorial inference: the same multitask loop could be applied at the level of detector channels or scalar descriptors extracted from spectra (coercive voltage, loop area), turning any multi-observation SPM protocol into an active-learning task space.
  • Editorial inference: a direct test of the transfer benefit would compare the joint model against two independent GPs on held-out co-located pairs; the paper reports a cross-task scatter of $r=0.304$ against the learned $\rho=0.752$, so the gap between raw and model-level correlation is where the transfer claim should be validated.
  • Editorial inference: the cost-aware formulation suggests a natural simulation benchmark: run the closed-loop policy on a pretrained multimodal dataset and compare total information gain per unit 'cost' against fixed-ratio mode allocation.
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 / 4 minor

Summary. The paper introduces multitask scanning probe microscopy, a closed-loop active-learning workflow in which a multitask Gaussian process with an intrinsic coregionalization model selects both the next measurement location and the next SPM protocol on an operating microscope. The method is demonstrated on a composition-spread AlScN wafer using tapping-mode and DART roughness as two tasks. Five paired seed measurements initialize the task-covariance matrix, after which 25 active-learning scans (17 model-selected, 8 forced-exploration) are executed with one mode per location. The authors report a learned cross-task correlation ρ=0.752, posterior mean maps that co-vary strongly (r=0.984), and an approximately 80% reduction in posterior standard deviation for both landscapes. The paper is transparent about several model-internal aspects of these numbers, noting in the Fig. 3c caption that the map agreement is partly imposed by the shared kernel and should not be read as independent validation.

Significance. If the central claim is supported—that measurements in one SPM mode can transfer information to another mode so that two response landscapes are reconstructed without measuring both modes everywhere—this would be a useful extension of autonomous SPM from spatial sampling to joint location-and-modality allocation, with clear relevance to wafer-scale and combinatorial characterization. The live execution of the loop (35 real scans, automated mode switching, stage motion) and the public availability of the code are concrete strengths. The paper is also unusually candid about the limitations of its own quantitative evidence, explicitly flagging that the posterior scatter is imposed by construction and that the five paired seeds are the primary constraint on ρ. The remaining weakness is that the load-bearing numbers—ρ=0.752, the 80% uncertainty reduction, and the cross-map agreement—are internal to the fitted model and are not checked against any held-out per-task ground truth or against an independent single-task baseline.

major comments (3)
  1. [Sec. III.E, Fig. 3] The central evidence for information transfer is model-internal. The learned ρ=0.752 is identified primarily from five co-located seed pairs, while the raw measured-versus-predicted cross-task scatter in Fig. 3b has r=0.304, and the posterior-mean scatter r=0.984 in Fig. 3c is, as the authors note, partly imposed by the shared spatial kernel. To support the claim in Sec. III.F that the experiment reconstructs two landscapes without measuring both modes at every position, the paper needs a held-out comparison against a single-task GP (or a diagonal-B ICM) trained on the same per-task observations. Reporting predictive RMSE and log-likelihood on unmeasured locations, or a leave-one-out analysis over the 35 scans, would provide the missing independent check.
  2. [Sec. III.F, Fig. 4f] The reported ~80% uncertainty reduction is the decrease in the model's own posterior standard deviation (0.176 nm and 0.235 nm versus prior values of 0.925 nm and 1.098 nm). This quantity depends directly on the fitted task covariance and does not measure actual prediction error. Because an overestimated ρ would produce overconfident transfer, the authors should report actual predictive error at held-out locations—ideally including a small set of paired measurements acquired after the loop—and show the same metric for the independent-task baseline. Without this, the headline uncertainty reduction does not by itself demonstrate a multitask advantage.
  3. [Sec. III.D and III.E] The uncertainty in the learned cross-task correlation is not quantified. With only five paired seed measurements constraining ρ, the point estimate 0.752 carries large uncertainty. The paper should report a confidence interval or posterior distribution for ρ (e.g., from a bootstrap over the five seed pairs or from the posterior over B), and should include a sensitivity analysis showing how the reconstructed maps and acquisition decisions change when ρ is fixed to smaller values such as 0.3 or 0.5. This is needed to establish that the transfer mechanism is identifiable from the data actually collected.
minor comments (4)
  1. [Sec. II.D, Eq. (12)] The annealing schedule β_eff = β0/(1+0.15 n_step) should specify whether n_step counts only model-driven active steps or also the forced-exploration steps; this affects reproducibility of the acquisition trajectory.
  2. [Sec. III.A] The thin-plate-spline height reference from 17 locations is a practical necessity; reporting the typical residual or error of this interpolant against a few validation points would help the reader judge whether height misestimation could bias the roughness measurements.
  3. [Sec. III.B] The roughness descriptor excludes pixels more than five standard deviations from the image mean before computing the standard deviation; for small images or high outlier densities this exclusion changes the estimator. State the image size and the typical number of excluded pixels.
  4. [Sec. III.E, Ref. 34] The claim that co-located measurements improve identifiability of the cross-task correlation and that non-overlapping designs recover it only slowly is central to the seed strategy, but Ref. 34 is an arXiv preprint. If the journal requires archival sources or if the claim is not yet peer-reviewed, the relevant analysis should be summarized in the text or an appendix.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the model's predictions are transparently fitted, and the paper explicitly disclaims the construction-imposed agreement as validation.

full rationale

The paper's central derivation is a standard multi-task Gaussian process with an intrinsic coregionalization model (Eqs. 3–5). The cross-task correlation ρ = B12 / (B11B22)^{1/2} = 0.752 is a model parameter fitted to the five paired seed measurements, and the paper explicitly states that this is 'a property of the fitted model rather than of the raw data' (Sec. III.E). The resulting posterior maps are used to demonstrate the closed-loop workflow, but the paper does not present the model-internal agreement (r = 0.984) as independent validation; it cautions that 'this agreement is partly imposed by construction and should not be read as independent validation' (Fig. 3c caption). The reported ~80% uncertainty reduction is a Bayesian posterior-variance computation, which is a legitimate model output, not a disguised prediction. No load-bearing self-citation chain or uniqueness theorem is invoked: the ICM model is cited to standard textbooks (Refs. 26, 27), and the identifiability comment (Ref. 34) is an external citation. The statistical fragility of estimating ρ from only five paired measurements is a robustness/correctness concern, not a circularity, because the paper does not claim that ρ is derived from the predictions it enables. The derivation chain is therefore self-contained relative to its stated assumptions, and the acknowledged construction-imposed agreement is not used as evidence.

Assumptions & free parameters 11 free parameters · 7 assumptions · 0 invented entities

The central demonstration rests on a small number of fitted model parameters (task covariance, kernel length scales, per-task noise, mean offsets) and on the domain assumption that the two roughness tasks share a common spatial correlation geometry. The cross-task correlation is identified from five paired seed measurements, and the acquisition policy contains several hand-chosen constants. No new physical entities are introduced.

free parameters (11)
  • Task-covariance matrix B (via W and kappa) = rho = 0.752; diagonal entries not quoted
    Fitted by maximum marginal likelihood on all 35 observations; determines how much information transfers between tapping and DART (Sec. III.E).
  • Spatial kernel length scales l_x, l_y = not quoted numerically (Fig. 4d)
    Fitted ARD length scales of the Matérn 5/2 kernel; set the spatial correlation range used in reconstruction.
  • Output scale sigma_f^2 = not quoted
    Fitted GP output variance; scales the posterior uncertainty statements.
  • Per-task noise sigma_t^2 = tapping 0.233 nm, DART 0.437 nm
    Fitted observation noise; the large difference is reported but not externally validated.
  • Per-task mean offsets mu_t = not quoted
    Fitted offsets account for systematic magnitude differences between modes (Sec. III.C).
  • Exploration annealing beta_0 and rate = beta_0 = 2.0, rate = 0.15
    Hand-chosen acquisition constants that control exploration-exploitation and affect which positions and modes are selected.
  • Forced-exploration interval = every 3 steps (8 of 25 active steps)
    Hand-chosen schedule; these random steps reduce the evidence for autonomous modality selection.
  • Edge mask parameters = soft suppression within 10 mm, hard exclusion within 3 mm of rim
    Hand-chosen penalties that suppress candidate positions near the wafer edge.
  • Roughness descriptor outlier threshold = 5 standard deviations
    Hand-chosen filter defining the scalar roughness from each image; changes the response values.
  • Seed selection random seed = 247
    Hand-chosen random seed for the five paired seed positions; small paired sample makes results potentially seed-dependent.
  • Rank R of task-covariance low-rank term = R = 2
    Hand-chosen rank of the shared latent representation in Eq. 4.
assumptions (7)
  • domain assumption Tapping and DART roughness share a common spatial correlation geometry (ICM factorization in Eq. 3-5)
    Sec. II.B states the tasks share the same kernel length scales; if false, cross-task information transfer is mis-specified.
  • domain assumption The scalar roughness descriptor (standard deviation of height after 5-sigma outlier exclusion) is a valid response for both tasks
    Sec. III.B defines the tasks through this descriptor; the learned correlation is only meaningful for this descriptor.
  • standard math Gaussian process prior with Matérn 5/2 kernel and additive homoskedastic per-task noise
    Standard GP assumptions from Ref. 33, used in Sec. II.B and III.C.
  • domain assumption Five paired seed measurements are sufficient to identify the cross-task correlation B
    Sec. III.E says the five co-located pairs are the primary constraint on rho; this is not supported by bootstrap or sensitivity analysis.
  • ad hoc to paper Random-scalarization UCB (Eq. 11) with the stated weights and annealing is a valid policy for joint location and task selection
    Sec. II.D introduces the policy; it is not benchmarked against single-task or cost-aware alternatives.
  • ad hoc to paper Forced exploration every third step prevents clustering and biased hyperparameter estimation
    Sec. II.D; the interval is chosen without quantitative support.
  • domain assumption Thin-plate spline interpolation of 17 height measurements gives adequate wafer-height compensation for stage motion
    Sec. III.A; if inaccurate, approaches or scans are mislocated, affecting data quality.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multitask Scanning Probe Microscopy." pith.science (2026). https://pith.science/paper/BHK7OMMT

@misc{pith2026260809104,
  author       = {Pith},
  title        = {Pith review of: Multitask Scanning Probe Microscopy},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BHK7OMMT}},
  note         = {Machine review of arXiv:2608.09104}
}
read the original abstract

Scanning probe microscopy provides nanoscale access to structural, electrical, electromechanical, magnetic, and mechanical properties of materials. Its increasing use for wafer-scale characterization and combinatorial materials exploration creates a need to distribute measurements efficiently across large spatial domains. This is particularly important when available modalities differ in acquisition time and potential for tip and sample damage, making exhaustive multimodal mapping over spatial grids impractical. Here, we demonstrate multitask scanning probe microscopy, a live, closed-loop workflow in which a multitask Gaussian process learns spatial and cross-modal relationships and autonomously selects both the next measurement location and the next experimental protocol. The approach is implemented on an automated large-sample atomic force microscope and demonstrated on a composition-spread AlScN wafer using tapping-mode and Dual AC Resonance Tracking (DART) measurements. Paired initial measurements establish the relation between the tasks, after which noncoincident measurements are used to update both response landscapes. The resulting workflow extends active learning in scanning probe microscopy from spatial sampling to autonomous allocation of measurement modalities and provides a basis for combining rapid, weakly perturbative imaging with slower contact, electrical, electromechanical, magnetic, or spectroscopic measurements.

Figures

Figures reproduced from arXiv: 2608.09104 by the authors.

Figure 3
Figure 3. Learned relationship between tapping and DART tasks. (a) Learned task-correlation matrix B, with off-diagonal entry ρ = B12/(B11B22)1/2 = 0.752, the learned correlation between the two measurement modes. (b) Cross-task scatter of measured roughness values; blue points [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Autonomous task allocation and reconstruction performance. (a) Total observations per task (16 tapping, 19 DART), split into the 5 paired seed observations and the 25 active-learning observations; of the latter, 17 were selected by the multitask Gaussian process and 8 were assigned by the forced-exploration rule. (b) Cumulative task observations over the 25 active-learning steps, excluding the seed phase. (c) Runnin… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 49 canonical work pages

  1. [1]

    K. Bian, C. Gerber, A. J. Heinrich, D. J. Müller, S. Scheuring and Y. Jiang, Nature Reviews Methods Primers 1 (1), 36 (2021)

    work page 2021

  2. [2]

    Berger, H

    R. Berger, H. J. Butt, M. B. Retschke and S. A. L. Weber, Macromolecular Rapid Communications 30 (14), 1167-1178 (2009)

    work page 2009

  3. [3]

    Nonnenmacher, M

    M. Nonnenmacher, M. P. O’Boyle and H. K. Wickramasinghe, Applied Physics Letters 58 (25), 2921-2923 (1991)

    work page 1991

  4. [4]

    Gruverman and S

    A. Gruverman and S. V. Kalinin, Journal of Materials Science 41 (1), 107-116 (2006)

    work page 2006

  5. [5]

    Kazakova, R

    O. Kazakova, R. Puttock, C. Barton, H. Corte-León, M. Jaafar, V. Neu and A. Asenjo, Journal of Applied Physics 125 (6), 060901 (2019)

    work page 2019

  6. [6]

    Martin, C

    Y. Martin, C. C. Williams and H. K. Wickramasinghe, Journal of Applied Physics 61 (10), 4723-4729 (1987)

    work page 1987

  7. [7]

    B. N. Slautin, Y. Liu, K. Barakati, Y. Liu, R. Emery, S. Hong, A. Dubey, V. V. Shvartsman, D. C. Lupascu, S. L. Sanchez, M. Ahmadi, Y. Kim, E. Strelcov, K. A. Brown, P. D. Rack and S. V. Kalinin, Applied Physics Reviews 12 (3), 031321 (2025)

    work page 2025

  8. [8]

    X. D. Xiang, X. Sun, G. Briceno, Y. Lou, K. A. Wang, H. Chang, W. G. Wallace- Freedman, S. W. Chen and P. G. Schultz, Science 268 (5218), 1738-1740 (1995)

    work page 1995

Show all 49 references
  1. [9]

    Chang, C

    H. Chang, C. Gao, I. Takeuchi, Y. Yoo, J. Wang, P. G. Schultz, X. D. Xiang, R. P. Sharma, M. Downes and T. Venkatesan, Applied Physics Letters 72 (17), 2185-2187 (1998)

    1998

  2. [10]

    R. A. Potyrailo, K. Rajan, K. Stoewe, I. Takeuchi, B. Chisholm and H. Lam, ACS Combinatorial Science 13 (6), 579-633 (2011)

    2011

  3. [11]

    Ludwig, npj Computational Materials 5 (1), 70 (2019)

    A. Ludwig, npj Computational Materials 5 (1), 70 (2019)

    2019

  4. [12]

    P. K. Hansma, J. P. Cleveland, M. Radmacher, D. A. Walters, P. E. Hillner, M. Bezanilla, M. Fritz, D. Vie, H. G. Hansma, C. B. Prater, J. Massie, L. Fukunaga, J. Gurley and V. Elings, Applied Physics Letters 64 (13), 1738-1740 (1994)

    1994

  5. [13]

    Jesse, A

    S. Jesse, A. P. Baddorf and S. V. Kalinin, Applied Physics Letters 88 (6), 062908 (2006)

    2006

  6. [14]

    J. Liu, J. K. Notbohm, R. W. Carpick and K. T. Turner, ACS Nano 4 (7), 3763-3772 (2010)

    2010

  7. [15]

    Raghavan, R

    A. Raghavan, R. Pant, I. Takeuchi, E. A. Eliseev, M. Checa, A. N. Morozovska, M. Ziatdinov, S. V. Kalinin and Y. Liu, ACS Nano 18 (37), 25591-25600 (2024)

    2024

  8. [16]

    S. V. Kalinin, M. Ziatdinov, J. Hinkle, S. Jesse, A. Ghosh, K. P. Kelley, A. R. Lupini, B. G. Sumpter and R. K. Vasudevan, ACS Nano 15 (8), 12604-12627 (2021)

    2021

  9. [17]

    R. K. Vasudevan, K. P. Kelley, J. Hinkle, H. Funakubo, S. Jesse, S. V. Kalinin and M. Ziatdinov, ACS Nano 15 (7), 11253-11262 (2021)

    2021

  10. [18]

    J. C. Thomas, A. Rossi, D. Smalley, L. Francaviglia, Z. Yu, T. Zhang, S. Kumari, J. A. Robinson, M. Terrones, M. Ishigami, E. Rotenberg, E. S. Barnard, A. Raja, E. Wong, D. F. Ogletree, M. M. Noack and A. Weber-Bargioni, npj Computational Materials 8 (1), 99 (2022)

    2022

  11. [19]

    Y. Liu, K. P. Kelley, R. K. Vasudevan, H. Funakubo, M. A. Ziatdinov and S. V. Kalinin, Nature Machine Intelligence 4 (4), 341-350 (2022)

    2022

  12. [20]

    Ziatdinov, Y

    M. Ziatdinov, Y. Liu, K. Kelley, R. Vasudevan and S. V. Kalinin, ACS Nano 16 (9), 13492-13512 (2022)

    2022

  13. [21]

    M. A. Ziatdinov, Y. Liu, A. N. Morozovska, E. A. Eliseev, X. Zhang, I. Takeuchi and S. V. Kalinin, Advanced Materials 34 (20), 2201345 (2022)

    2022

  14. [22]

    Krull, P

    A. Krull, P. Hirsch, C. Rother, A. Schiffrin and C. Krull, Communications Physics 3 (1), 54 (2020)

    2020

  15. [23]

    Y. Liu, U. Pratiush, J. Bemis, R. Proksch, R. Emery, P. D. Rack, Y. C. Liu, J. C. Yang, S. Udovenko, S. Trolier-McKinstry and S. V. Kalinin, Review of Scientific Instruments 95 (9), 093701 (2024)

    2024

  16. [24]

    Y. Liu, R. K. Vasudevan, K. P. Kelley, H. Funakubo, M. Ziatdinov and S. V. Kalinin, npj Computational Materials 9 (1), 34 (2023)

    2023

  17. [25]

    B. N. Slautin, U. Pratiush, I. N. Ivanov, Y. Liu, R. Pant, X. Zhang, I. Takeuchi, M. A. Ziatdinov and S. V. Kalinin, Digital Discovery 3 (8), 1602-1611 (2024)

    2024

  18. [26]

    E. V. Bonilla, K. M. A. Chai and C. K. I. Williams, in Advances in Neural Information Processing Systems (Curran Associates, 2008), Vol. 20, pp. 153-160

    2008

  19. [27]

    M. A. Álvarez, L. Rosasco and N. D. Lawrence, Foundations and Trends in Machine Learning 4 (3), 195-266 (2012)

    2012

  20. [28]

    Akiyama, T

    M. Akiyama, T. Kamohara, K. Kano, A. Teshigahara, Y. Takeuchi and N. Kawahara, Advanced Materials 21 (5), 593-596 (2009)

    2009

  21. [29]

    Fichtner, N

    S. Fichtner, N. Wolff, F. Lofink, L. Kienle and B. Wagner, Journal of Applied Physics 125 (11), 114103 (2019)

    2019

  22. [30]

    B. J. Rodriguez, C. Callahan, S. V. Kalinin and R. Proksch, Nanotechnology 18 (47), 475504 (2007)

    2007

  23. [31]

    Gannepalli, D

    A. Gannepalli, D. G. Yablon, A. H. Tsou and R. Proksch, Nanotechnology 22 (35), 355705 (2011)

    2011

  24. [32]

    Bradler, A

    S. Bradler, A. Schirmeisen and B. Roling, Journal of Applied Physics 123 (3), 035106 (2018)

    2018

  25. [33]

    C. E. Rasmussen and C. K. I. Williams, Gaussian Processes for Machine Learning. (MIT Press, Cambridge, MA, 2006)

    2006

  26. [34]

    Pitfalls and Remedies for Multi- Task Bayesian Optimization,

    C. Hvarfner, S. Daulton, M. Balandat and E. Bakshy, “Pitfalls and Remedies for Multi- Task Bayesian Optimization,” arXiv:2607.09073 (2026)

    2026 arXiv

  27. [35]

    Snoek, H

    J. Snoek, H. Larochelle and R. P. Adams, in Advances in Neural Information Processing Systems (Curran Associates, 2012), Vol. 25

    2012

  28. [36]

    Ghosh, S

    A. Ghosh, S. V. Kalinin and M. A. Ziatdinov, APL Machine Learning 1 (4), 046102 (2023)

    2023

  29. [37]

    Y. Liu, A. N. Morozovska, E. A. Eliseev, K. P. Kelley, R. Vasudevan, M. Ziatdinov and S. V. Kalinin, Patterns 4 (3), 100704 (2023)

    2023

  30. [38]

    M. C. Kennedy and A. O’Hagan, Biometrika 87 (1), 1-13 (2000)

    2000

  31. [39]

    M. A. Ziatdinov, A. Ghosh and S. V. Kalinin, Machine Learning: Science and Technology 3 (1), 015003 (2022)

    2022

  32. [40]

    S. L. Sanchez, E. Foadian, M. Ziatdinov, J. Yang, S. V. Kalinin, Y. Liu and M. Ahmadi, Digital Discovery 3 (8), 1577-1590 (2024)

    2024

  33. [41]

    Swersky, J

    K. Swersky, J. Snoek and R. P. Adams, in Advances in Neural Information Processing Systems (Curran Associates, 2013), Vol. 26

    2013

  34. [42]

    Knowles, IEEE Transactions on Evolutionary Computation 10 (1), 50-66 (2006)

    J. Knowles, IEEE Transactions on Evolutionary Computation 10 (1), 50-66 (2006)

    2006

  35. [43]

    Paria, K

    B. Paria, K. Kandasamy and B. Póczos, presented at the Proceedings of the 35th Conference on Uncertainty in Artificial Intelligence (UAI), Proceedings of Machine Learning Research, 2020

    2020

  36. [44]

    Srinivas, A

    N. Srinivas, A. Krause, S. M. Kakade and M. Seeger, presented at the Proceedings of the 27th International Conference on Machine Learning (ICML), 2010

    2010

  37. [45]

    R. W. Kennard and L. A. Stone, Technometrics 11 (1), 137-148 (1969)

    1969

  38. [46]

    Takeno, H

    S. Takeno, H. Fukuoka, Y. Tsukada, T. Koyama, M. Shiga, I. Takeuchi and M. Karasuyama, presented at the Proceedings of the 37th International Conference on Machine Learning (ICML), Proceedings of Machine Learning Research, 2020

    2020

  39. [47]

    Duchon, in Constructive Theory of Functions of Several Variables (Springer, Berlin, 1977), Vol

    J. Duchon, in Constructive Theory of Functions of Several Variables (Springer, Berlin, 1977), Vol. 571, pp. 85-100

    1977

  40. [48]

    D. P. Kingma and J. Ba, presented at the Proceedings of the 3rd International Conference on Learning Representations (ICLR), 2015

    2015

  41. [49]

    J. R. Gardner, G. Pleiss, D. Bindel, K. Q. Weinberger and A. G. Wilson, in Advances in Neural Information Processing Systems (Curran Associates, 2018), Vol. 31

    2018

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.