pith. sign in

arxiv: 2603.20959 · v2 · submitted 2026-03-21 · 📊 stat.CO · math.PR

Surrogate-Guided Adaptive Importance Sampling for Failure Probability Estimation

Ashwin Renganathan , Annie S. Booth This is my paper

Pith reviewed 2026-05-15 06:41 UTC · model grok-4.3

classification 📊 stat.CO math.PR
keywords importance samplingGaussian processkernel density estimationfailure probabilitysurrogate modeladaptive samplingrare events
0
0 comments X

The pith

KDE-AIS converges asymptotically to the optimal zero-variance importance sampling density while estimating failure probabilities efficiently from shared surrogate evaluations.

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

The paper introduces a single-stage method for estimating small failure probabilities using expensive simulations of a limit state function. Instead of first building a surrogate and then sampling separately, it trains both the Gaussian process surrogate and an adaptive importance sampling proposal from the same evaluations. Kernel density estimation is used to construct the proposal density adaptively. The authors prove that this KDE-AIS density converges to the ideal zero-variance sampler in total variation. Empirical results show it achieves accurate estimates with fewer oracle calls than previous two-stage or adaptive methods.

Core claim

KDE-AIS trains a Gaussian process surrogate for the limit state function and uses kernel density estimation on the same evaluations to adaptively build the importance sampling proposal. This leads to the proposal density converging asymptotically to the optimal zero-variance importance sampling density in total variation. The approach enables accurate estimation of small failure probabilities with a limited budget of expensive oracle evaluations, outperforming state-of-the-art Gaussian process based adaptive importance sampling techniques.

What carries the argument

KDE-AIS, which integrates Gaussian process surrogates with kernel density estimation to adaptively construct the IS proposal density from shared oracle evaluations of the limit state.

If this is right

  • The single-stage approach uses a shared budget more efficiently than conventional two-stage methods.
  • KDE-AIS density converges to the optimal zero-variance IS density in total variation.
  • Small failure probabilities can be estimated accurately with relatively few oracle evaluations.
  • Empirical performance exceeds previous Gaussian process based adaptive importance sampling methods.

Where Pith is reading between the lines

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

  • If the convergence holds in practice, it could enable reliable analysis of rare events in high-stakes systems like aerospace or nuclear engineering with reduced simulation budgets.
  • Future work might test the method on problems where the limit state is less smooth or in very high dimensions.
  • The approach could be combined with other surrogate types beyond Gaussian processes for broader applicability.

Load-bearing premise

The limit state function must be sufficiently smooth for the Gaussian process surrogate to reliably guide the construction of the adaptive proposal density.

What would settle it

If the estimated proposal density fails to approach the true optimal importance sampling density in total variation distance for a smooth but high-dimensional limit state function, or if the failure probability estimates remain inaccurate despite increasing the number of evaluations.

Figures

Figures reproduced from arXiv: 2603.20959 by Annie S. Booth, Ashwin Renganathan.

Figure 1
Figure 1. Figure 1: Overview of the proposed “single-stage” approach where high fidelity (HF) oracle [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: KDE-AIS on the Herbie function (a) and four branch function (b). Interior top left: true function with failure boundaries as red lines. Interior top right: predicted surrogate failure region after n = 510. Interior bottom left/right: optimal/predicted (after n = 510) IS proposal densities. the total samples. In methods such as GPAIS, the lack of an automatic means to densely sample X can potentially miss i… view at source ↗
Figure 3
Figure 3. Figure 3: Snapshots of the posterior GP mean (µn) for the Herbie function (t = 2.0). Red lines represent learned limit sate gb = t; black circles are n = 5 seed points; white dots are samples drawn from the proposal qn. 17 [PITH_FULL_IMAGE:figures/full_fig_p017_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Snapshots of the posterior GP mean (µn) for the Four branch function (t = 2.0). Red lines represent learned limit sate gb = t; black circles are n = 5 seed points; white dots are samples drawn from the proposal qn. 18 [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Evolution of PbF against the number of oracle evaluations for the synthetic experi￾ments. Since the two-stage procedures are not sequential, their variability across repetitions is indicated by violin plots at the number of evaluations used for surrogate training. Shaded regions indicate [min, max] range. limit-state function g : R 2 → R is defined as g(x) = min  g1(x), g2(x), g3(x), g4(x) [PITH_FULL_IMA… view at source ↗
Figure 6
Figure 6. Figure 6: Evolution of PbF against the number of oracle evaluations for the synthetic experi￾ments. Shaded regions indicate [min, max] range. 5.2.1 Cantilever beam Next, we consider a prismatic cantilever beam under end loads, where the maximum deflection of the beam under the load is used to assess failure. The input vector is x = (P, L, E, Θ)⊤ ∈ R 4 , where P is the end load, L is the span, E is the Young’s modulu… view at source ↗
read the original abstract

We consider the sample efficient estimation of failure probabilities from expensive oracle evaluations of a limit state function via importance sampling (IS). In contrast to conventional ``two stage'' approaches, which first train a surrogate model for the limit state and then construct an IS proposal to estimate failure probability using separate oracle evaluations, we propose a \emph{single stage} approach where a Gaussian process surrogate and a surrogate for the optimal (zero-variance) IS density are trained from shared evaluations of the oracle, making better use of a limited budget. With such an approach, small failure probabilities can be learned with relatively few oracle evaluations. We propose \emph{kernel density estimation adaptive importance sampling} (\texttt{KDE-AIS}), which combines Gaussian process surrogates with kernel density estimation to adaptively construct the IS proposal density, leading to sample efficient estimation of failure probabilities. We show that \texttt{KDE-AIS} density asymptotically converges to the optimal zero-variance IS density in total variation. Empirically, \texttt{KDE-AIS} enables accurate and sample efficient estimation of failure probabilities compared to the state of the art, including previous work on Gaussian process based adaptive importance sampling.

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 paper proposes KDE-AIS, a single-stage method that jointly trains a Gaussian process surrogate for the limit state function and uses kernel density estimation on the same oracle evaluations to adaptively construct an importance sampling proposal density. It claims that this KDE-AIS proposal converges in total variation to the optimal zero-variance IS density and demonstrates empirical gains in sample efficiency and accuracy for estimating small failure probabilities relative to two-stage surrogate-IS baselines and prior GP-based adaptive IS methods.

Significance. If the total-variation convergence result can be rigorously established despite the adaptive dependence, the work would offer a principled way to integrate surrogate modeling with adaptive importance sampling, potentially reducing the number of expensive limit-state evaluations needed for rare-event estimation in reliability analysis. The empirical results, if robust, indicate practical improvements over existing approaches.

major comments (3)
  1. [Abstract and theoretical analysis] The asymptotic total-variation convergence of the KDE-AIS density to the optimal zero-variance IS density (abstract and theoretical section): standard KDE consistency results require i.i.d. samples, yet the adaptive construction draws successive samples from an evolving proposal that depends on the sequentially updated GP surrogate. Without an explicit ergodicity, martingale, or mixing argument to control the induced dependence, the claimed convergence does not necessarily follow from existing KDE theory.
  2. [Method description] The weakest assumption that the limit state function is sufficiently smooth for the GP surrogate to reliably guide the adaptive proposal (implicit in the method definition): the paper should state explicit conditions on the kernel and the function class (e.g., Hölder or Sobolev regularity) under which the GP posterior concentrates sufficiently to ensure the KDE step targets the correct region, and discuss failure modes when these are violated.
  3. [Experiments] Empirical comparisons (experimental section): the reported accuracy and efficiency gains versus prior GP-AIS methods should be accompanied by sensitivity checks on the free parameters (GP kernel hyperparameters and KDE bandwidth) to confirm that the advantages are not artifacts of post-hoc tuning or favorable data selection.
minor comments (2)
  1. Notation for the adaptive proposal density could be made more explicit to distinguish the surrogate-guided KDE estimate from the target optimal density throughout the derivations.
  2. Figure captions and axis labels in the numerical results should include the specific failure probability values and sample budgets used to facilitate direct comparison with the claimed efficiency gains.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments. We address each major point below and describe the revisions that will be incorporated into the next version of the manuscript.

read point-by-point responses

  1. Referee: [Abstract and theoretical analysis] The asymptotic total-variation convergence of the KDE-AIS density to the optimal zero-variance IS density (abstract and theoretical section): standard KDE consistency results require i.i.d. samples, yet the adaptive construction draws successive samples from an evolving proposal that depends on the sequentially updated GP surrogate. Without an explicit ergodicity, martingale, or mixing argument to control the induced dependence, the claimed convergence does not necessarily follow from existing KDE theory.

    Authors: We agree that the adaptive dependence requires explicit justification. In the revision we will expand the theoretical analysis to include a proof that first establishes almost-sure convergence of the adaptive proposal to the zero-variance optimum (via GP posterior consistency) and then invokes ergodic theorems for the resulting Markov chain to obtain total-variation convergence of the KDE estimator. The argument will be stated under the same regularity conditions already used for the GP. revision: yes

  2. Referee: [Method description] The weakest assumption that the limit state function is sufficiently smooth for the GP surrogate to reliably guide the adaptive proposal (implicit in the method definition): the paper should state explicit conditions on the kernel and the function class (e.g., Hölder or Sobolev regularity) under which the GP posterior concentrates sufficiently to ensure the KDE step targets the correct region, and discuss failure modes when these are violated.

    Authors: We will add an explicit assumptions subsection stating that the limit-state function lies in a Sobolev space of order s > d/2 and that the GP uses a Matérn kernel with smoothness parameter ν ≥ s. Under these conditions the posterior concentrates at the rate needed for the KDE to target the failure region. We will also include a short discussion of failure modes (e.g., discontinuous or lower-regularity functions) that can produce surrogate bias and degrade the IS estimator. revision: yes

  3. Referee: [Experiments] Empirical comparisons (experimental section): the reported accuracy and efficiency gains versus prior GP-AIS methods should be accompanied by sensitivity checks on the free parameters (GP kernel hyperparameters and KDE bandwidth) to confirm that the advantages are not artifacts of post-hoc tuning or favorable data selection.

    Authors: We will augment the experimental section with sensitivity studies that vary the GP length-scale and variance (around MLE values) and the KDE bandwidth (Silverman’s rule versus cross-validation). The additional tables and figures will demonstrate that the reported gains in accuracy and sample efficiency remain stable across these choices. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation relies on external KDE consistency applied to adaptive process

full rationale

The paper constructs KDE-AIS by training a GP surrogate on oracle evaluations and using KDE on the resulting adaptive samples to build the IS proposal. The central claim is an asymptotic total-variation convergence result to the zero-variance optimum. This is presented as a theorem whose proof invokes standard KDE consistency arguments rather than re-deriving the optimum from the fitted GP or KDE parameters themselves. No equation reduces the claimed convergence to a fitted parameter by construction, no self-citation supplies a uniqueness theorem that forces the result, and the adaptive dependence is addressed (or assumed away) via external ergodicity conditions rather than being smuggled in as a definition. The method is therefore self-contained against external benchmarks and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

The approach rests on standard assumptions for Gaussian process regression and kernel density estimation plus the existence of an optimal zero-variance importance density; no new entities are postulated.

free parameters (2)
  • GP kernel hyperparameters
    Length-scale and variance parameters of the Gaussian process surrogate are fitted to the shared oracle data.
  • KDE bandwidth
    Bandwidth parameter controlling the smoothness of the kernel density estimate used for the adaptive proposal.
axioms (1)
  • domain assumption The limit state function is continuous and the input domain permits a well-defined optimal importance density.
    Invoked to guarantee that the KDE-AIS proposal can converge to the zero-variance optimum.

pith-pipeline@v0.9.0 · 5502 in / 1297 out tokens · 21432 ms · 2026-05-15T06:41:45.714977+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

72 extracted references · 72 canonical work pages · 1 internal anchor

  1. [1]

    A. H-S. Ang and W. H. (see also Ang et al.) Tang. Optimal importance-sampling density estimator.Journal of Engineering Mechanics, 118(6):1146–1164, 1992. doi: 10.1061/(ASCE)0733-9399(1992)118:6(1146)

    work page doi:10.1061/(asce)0733-9399(1992)118:6(1146 1992

  2. [2]

    Siu-Kui Au and James L. Beck. A new adaptive importance sampling scheme for reliability evaluation.Structural Safety, 21(2):135–158, 1999. doi: 10.1016/S0167-4730(99)00014-4

    work page doi:10.1016/s0167-4730(99)00014-4 1999

  3. [3]

    Siu-Kui Au and James L. Beck. Estimation of small failure probabilities in high dimensions by subset simulation.Probabilistic Engineering Mechanics, 16(4):263–277,

    work page

  4. [4]

    doi: 10.1016/S0266-8920(01)00019-4

    work page doi:10.1016/s0266-8920(01)00019-4

  5. [5]

    Adaptive design of experiments for conservative estimation of excursion sets.Techno- metrics, 63(1):13–26, 2021

    Dario Azzimonti, David Ginsbourger, Cl´ ement Chevalier, Julien Bect, and Yann Richet. Adaptive design of experiments for conservative estimation of excursion sets.Techno- metrics, 63(1):13–26, 2021

    work page 2021

  6. [6]

    Kriging-based adaptive importance sampling algorithms for rare event estimation.Structural Safety, 44:1–10, 2013

    Mathieu Balesdent, J´ erˆ ome Morio, and Julien Marzat. Kriging-based adaptive importance sampling algorithms for rare event estimation.Structural Safety, 44:1–10, 2013. doi: 10.1016/j.strusafe.2013.05.002

    work page doi:10.1016/j.strusafe.2013.05.002 2013

  7. [7]

    Sequential design of computer experiments for the estimation of a probability of failure

    Julien Bect, David Ginsbourger, Ling Li, Victor Picheny, and Emmanuel Vazquez. Sequential design of computer experiments for the estimation of a probability of failure. Statistics and Computing, 22(3):773–793, 2012

    work page 2012

  8. [8]

    Bayesian subset simulation.SIAM/ASA Journal on Uncertainty Quantification, 5(1):762–786, 2017

    Julien Bect, Ling Li, and Emmanuel Vazquez. Bayesian subset simulation.SIAM/ASA Journal on Uncertainty Quantification, 5(1):762–786, 2017. doi: 10.1137/16M1078276

    work page doi:10.1137/16m1078276 2017

  9. [9]

    Efficient global reliability analysis for nonlinear implicit performance functions.AIAA journal, 46(10):2459–2468, 2008

    Barron J Bichon, Michael S Eldred, Laura Painton Swiler, Sandaran Mahadevan, and John M McFarland. Efficient global reliability analysis for nonlinear implicit performance functions.AIAA journal, 46(10):2459–2468, 2008

    work page 2008

  10. [10]

    Two-stage design for failure probability estimation with gaussian process surrogates.Journal of Quality Technology, pages 1–17, 2025

    Annie S Booth and S Ashwin Renganathan. Two-stage design for failure probability estimation with gaussian process surrogates.Journal of Quality Technology, pages 1–17, 2025

    work page 2025

  11. [11]

    Actively learning deep gaus- sian process models for failure contour and probability estimation

    Annie S Booth, Robert Gramacy, and Ashwin Renganathan. Actively learning deep gaus- sian process models for failure contour and probability estimation. InAIAA SCITECH 2024 Forum, page 0577, 2024

    work page 2024

  12. [12]

    Contour location for reliability in airfoil simulation experiments using deep gaussian processes.The Annals of Applied Statistics, 19(1):191–211, 2025

    Annie S Booth, S Ashwin Renganathan, and Robert B Gramacy. Contour location for reliability in airfoil simulation experiments using deep gaussian processes.The Annals of Applied Statistics, 19(1):191–211, 2025

    work page 2025

  13. [13]

    Z. I. Botev, J. F. Grotowski, and D. P. Kroese. Kernel density estimation via diffusion. Ann. Statist., 38(5):2916–2957, 2010

    work page 2010

  14. [14]

    Nonparametric density estimation for stratified samples.Statistics & Probability Letters, 78(14):2194–2200, 2008

    Robert Breunig. Nonparametric density estimation for stratified samples.Statistics & Probability Letters, 78(14):2194–2200, 2008. doi: 10.1016/j.spl.2008.01.099. 30

    work page doi:10.1016/j.spl.2008.01.099 2008

  15. [15]

    Adaptive importance sampling: The past, the present, and the future

    Monica F Bugallo, Victor Elvira, Luca Martino, David Luengo, Joaquin Miguez, and Petar M Djuric. Adaptive importance sampling: The past, the present, and the future. IEEE Signal Processing Magazine, 34(4):60–79, 2017

    work page 2017

  16. [16]

    Buskirk and Sharon L

    Trent D. Buskirk and Sharon L. Lohr. Asymptotic properties of kernel density estimation with complex survey data.Journal of Statistical Planning and Inference, 128(1):165–190,

    work page

  17. [17]

    doi: 10.1016/j.jspi.2003.09.036

    work page doi:10.1016/j.jspi.2003.09.036 2003

  18. [18]

    An improved adaptive kriging-based importance technique for sampling multiple failure regions of low probability.Reliability Engineering & System Safety, 131:109–117, 2014

    Francesco Cadini, Agnese Santos, and Enrico Zio. An improved adaptive kriging-based importance technique for sampling multiple failure regions of low probability.Reliability Engineering & System Safety, 131:109–117, 2014. doi: 10.1016/j.ress.2014.06.023

    work page doi:10.1016/j.ress.2014.06.023 2014

  19. [19]

    Fast parallel kriging-based stepwise uncertainty reduction with application to the identification of an excursion set.Technometrics, 56(4):455–465, 2014

    Cl´ ement Chevalier, Julien Bect, David Ginsbourger, Emmanuel Vazquez, Victor Picheny, and Yann Richet. Fast parallel kriging-based stepwise uncertainty reduction with application to the identification of an excursion set.Technometrics, 56(4):455–465, 2014

    work page 2014

  20. [20]

    Entropy-based adaptive design for contour finding and estimating reliability.arXiv preprint arXiv:2105.11357, 2021

    D Austin Cole, Robert B Gramacy, James E Warner, Geoffrey F Bomarito, Patrick E Leser, and William P Leser. Entropy-based adaptive design for contour finding and estimating reliability.arXiv preprint arXiv:2105.11357, 2021

    work page arXiv 2021

  21. [21]

    Dalbey and Laura P

    Keith R. Dalbey and Laura P. Swiler. Gaussian process adaptive importance sampling. International Journal for Uncertainty Quantification, 4(2), 2014. doi: 10.1615/Int.J. UncertaintyQuantification.2013006330

    work page doi:10.1615/int.j 2014

  22. [22]

    REIN: Reliability estimation via importance sampling with normalizing flows.Reliability Engineering & System Safety, 242:109729,

    Agnimitra Dasgupta and Erik Johnson. REIN: Reliability estimation via importance sampling with normalizing flows.Reliability Engineering & System Safety, 242:109729,

    work page

  23. [23]

    doi: 10.1016/j.ress.2023.109729

    work page doi:10.1016/j.ress.2023.109729 2023

  24. [24]

    The geometry of random vibrations and solutions by form and sorm

    A Der Kiureghian. The geometry of random vibrations and solutions by form and sorm. Probabilistic Engineering Mechanics, 15(1):81–90, 2000

    work page 2000

  25. [25]

    Metamodel-based importance sampling for structural reliability analysis.Probabilistic Engineering Mechanics, 33: 47–57, 2013

    Vincent Dubourg, Franck Deheeger, and Bruno Sudret. Metamodel-based importance sampling for structural reliability analysis.Probabilistic Engineering Mechanics, 33: 47–57, 2013. doi: 10.1016/j.probengmech.2013.02.002

    work page doi:10.1016/j.probengmech.2013.02.002 2013

  26. [26]

    A sur version of the bichon criterion for excursion set estimation

    Cl´ ement Duhamel, C´ eline Helbert, Miguel Munoz Zuniga, Cl´ ementine Prieur, and Delphine Sinoquet. A sur version of the bichon criterion for excursion set estimation. Statistics and Computing, 33(2):41, 2023

    work page 2023

  27. [27]

    AK-MCS: An active learning reliability method combining kriging and monte carlo simulation.Structural Safety, 33 (2):145–154, 2011

    Benjamin Echard, Nicolas Gayton, and Michel Lemaire. AK-MCS: An active learning reliability method combining kriging and monte carlo simulation.Structural Safety, 33 (2):145–154, 2011. doi: 10.1016/j.strusafe.2011.01.002

    work page doi:10.1016/j.strusafe.2011.01.002 2011

  28. [28]

    Elvira, L

    V. Elvira, L. Martino, D. Luengo, and M. F. Bugallo. Generalized multiple importance sampling.arXiv:1511.03095, 2019

    work page arXiv 2019

  29. [29]

    Non-parametric estimation of a multivariate probability density

    Vassiliy A Epanechnikov. Non-parametric estimation of a multivariate probability density. Theory of Probability & Its Applications, 14(1):153–158, 1969. 31

    work page 1969

  30. [30]

    Weighted uniform consistency of kernel density estimators.The Annals of Probability, 32(3B):2570–2605, 2004

    Evarist Gine, Vladimir Koltchinskii, and Joel Zinn. Weighted uniform consistency of kernel density estimators.The Annals of Probability, 32(3B):2570–2605, 2004. doi: 10.1214/009117904000000063

    work page doi:10.1214/009117904000000063 2004

  31. [31]

    Active learning for level set estimation

    Alkis Gotovos, Nathalie Casati, Gregory Hitz, and Andreas Krause. Active learning for level set estimation. InTwenty-Third International Joint Conference on Artificial Intelligence, 2013

    work page 2013

  32. [32]

    Gramacy.Surrogates: Gaussian Process Modeling, Design and Optimization for the Applied Sciences

    Robert B. Gramacy.Surrogates: Gaussian Process Modeling, Design and Optimization for the Applied Sciences. Chapman Hall/CRC, Boca Raton, Florida, 2020. URL http://bobby.gramacy.com/surrogates

    work page 2020

  33. [33]

    Second-order reliability methods: a review and comparative study.Structural and multidisciplinary optimization, 64(6):3233–3263, 2021

    Zhangli Hu, Rami Mansour, M˚ arten Olsson, and Xiaoping Du. Second-order reliability methods: a review and comparative study.Structural and multidisciplinary optimization, 64(6):3233–3263, 2021

    work page 2021

  34. [34]

    Overview of structural reliability analysis methods—part i: Local reliability methods.Incertitudes et fiabilit´ e des syst` emes multiphysiques, 17(1):1–10, 2017

    Changwu Huang, Abdelkhalak El Hami, and Boucha¨ ıb Radi. Overview of structural reliability analysis methods—part i: Local reliability methods.Incertitudes et fiabilit´ e des syst` emes multiphysiques, 17(1):1–10, 2017

    work page 2017

  35. [35]

    Reliability analysis of slope stability under seismic condition during a given exposure time.Landslides, 15 (11):2303–2313, 2018

    HW Huang, SC Wen, J Zhang, FY Chen, JR Martin, and H Wang. Reliability analysis of slope stability under seismic condition during a given exposure time.Landslides, 15 (11):2303–2313, 2018

    work page 2018

  36. [36]

    Assessing small failure probabilities by ak–ss: An active learning method combining kriging and subset simulation.Structural Safety, 59: 86–95, 2016

    Xianfeng Huang, Jie Chen, and Su Li. Assessing small failure probabilities by ak–ss: An active learning method combining kriging and subset simulation.Structural Safety, 59: 86–95, 2016. doi: 10.1016/j.strusafe.2015.12.003

    work page doi:10.1016/j.strusafe.2015.12.003 2016

  37. [37]

    Efficient Global Optimization of Expensive Black-Box Functions

    Donald R. Jones, Matthias Schonlau, and William J. Welch. Efficient global optimization of expensive black-box functions.Journal of Global Optimization, 13(4):455–492, 1998. doi: 10.1023/A:1008306431147

    work page doi:10.1023/a:1008306431147 1998

  38. [38]

    H. K. H. Lee, R. B. Gramacy, C. Linkletter, and G. A. Gray. Optimization subject to hidden constraints via statistical emulation.Pacific Journal of Optimization, 7(3): 467–478, 2011

    work page 2011

  39. [39]

    Soon Man Lee, Jae-Hun Song, and Nam-Ho Kim. An adaptive importance sampling method with a kriging model and kernel sampling density for structural reliability analysis.Journal of Mechanical Science and Technology, 31:5873–5882, 2017. doi: 10.1007/s12206-017-1119-8

    work page doi:10.1007/s12206-017-1119-8 2017

  40. [40]

    An efficient surrogate-based method for computing rare failure probability.Journal of Computational Physics, 230(24):8683–8697, 2011

    Jing Li, Jinglai Li, and Dongbin Xiu. An efficient surrogate-based method for computing rare failure probability.Journal of Computational Physics, 230(24):8683–8697, 2011

    work page 2011

  41. [41]

    Nonparametric importance sampling for wind turbine reliability analysis.Annals of Applied Statistics, 15(4):1850–1873, 2021

    Shiyu Li, Yi Peng, and Eui-Young Byon. Nonparametric importance sampling for wind turbine reliability analysis.Annals of Applied Statistics, 15(4):1850–1873, 2021. doi: 10.1214/20-AOAS1438. 32

    work page doi:10.1214/20-aoas1438 2021

  42. [42]

    Credible occurrence probabilities for extreme geophysical events: Earth- quakes, volcanic eruptions, magnetic storms.Geophysical Research Letters, 39(10), 2012

    Jeffrey J Love. Credible occurrence probabilities for extreme geophysical events: Earth- quakes, volcanic eruptions, magnetic storms.Geophysical Research Letters, 39(10), 2012

    work page 2012

  43. [43]

    Courier Corporation, 2006

    Henrik O Madsen, Steen Krenk, and Niels Christian Lind.Methods of structural safety. Courier Corporation, 2006

    work page 2006

  44. [44]

    Contour location via entropy reduction leveraging multiple information sources

    Alexandre N Marques, Remi R Lam, and Karen E Willcox. Contour location via entropy reduction leveraging multiple information sources.arXiv preprint arXiv:1805.07489, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  45. [45]

    Statistical methods for extreme event attribution in climate science.Annual Review of Statistics and Its Application, 7 (1):89–110, 2020

    Philippe Naveau, Alexis Hannart, and Aur´ elien Ribes. Statistical methods for extreme event attribution in climate science.Annual Review of Statistics and Its Application, 7 (1):89–110, 2020

    work page 2020

  46. [46]

    Reliability approximation for solid shaft under gamma setup.Journal of Reliability and Statistical Studies, pages 11–17, 2014

    Sadananda Nayek, Babulal Seal, and Dilip Roy. Reliability approximation for solid shaft under gamma setup.Journal of Reliability and Statistical Studies, pages 11–17, 2014

    work page 2014

  47. [47]

    Adaptive importance sampling in monte carlo integration.Journal of statistical computation and simulation, 41(3-4):143–168, 1992

    Man-Suk Oh and James O Berger. Adaptive importance sampling in monte carlo integration.Journal of statistical computation and simulation, 41(3-4):143–168, 1992

    work page 1992

  48. [48]

    A. B. Owen.Monte Carlo Theory, Methods and Examples. Stanford University, 2013

    work page 2013

  49. [49]

    E. Parzen. On estimation of a probability density function and mode.Ann. Math. Statist., 33(3):1065–1076, 1962

    work page 1962

  50. [50]

    Multifidelity importance sampling.Computer Methods in Applied Mechanics and Engineering, 300: 490–509, 2016

    Benjamin Peherstorfer, Tiangang Cui, Youssef Marzouk, and Karen Willcox. Multifidelity importance sampling.Computer Methods in Applied Mechanics and Engineering, 300: 490–509, 2016

    work page 2016

  51. [51]

    Variance reduction using multiple importance sampling with adaptive kriging (ak-amis).PhD Thesis / preprint, University of Liverpool, 2023

    Augustin Persoons, Matteo Broggi, and Michael Beer. Variance reduction using multiple importance sampling with adaptive kriging (ak-amis).PhD Thesis / preprint, University of Liverpool, 2023. Available as: https://livrepository.liverpool.ac.uk/3171699/ 1/APersoons.pdf

    work page arXiv 2023

  52. [52]

    Adaptive designs of experiments for accurate approximation of a target region

    Victor Picheny, David Ginsbourger, Olivier Roustant, Raphael T Haftka, and Nam-Ho Kim. Adaptive designs of experiments for accurate approximation of a target region. 2010

    work page 2010

  53. [53]

    Sequential experiment design for contour estimation from complex computer codes.Technometrics, 50(4):527–541, 2008

    Pritam Ranjan, Derek Bingham, and George Michailidis. Sequential experiment design for contour estimation from complex computer codes.Technometrics, 50(4):527–541, 2008

    work page 2008

  54. [54]

    C. E. Rasmussen and C. K. I. Williams.Gaussian Processes for Machine Learning. MIT Press, 2006. doi: 10.7551/mitpress/3206.001.0001

    work page doi:10.7551/mitpress/3206.001.0001 2006

  55. [55]

    Camera: A method for cost-aware, adaptive, multifidelity, efficient reliability analysis.Journal of Computational Physics, 472:111698, 2023

    S Ashwin Renganathan, Vishwas Rao, and Ionel M Navon. Camera: A method for cost-aware, adaptive, multifidelity, efficient reliability analysis.Journal of Computational Physics, 472:111698, 2023. 33

    work page 2023

  56. [56]

    Gaussian mixture models

    Douglas Reynolds. Gaussian mixture models. InEncyclopedia of biometrics, pages 827–832. Springer, 2015

    work page 2015

  57. [57]

    Romero, Laura P

    Vicente J. Romero, Laura P. Swiler, Mohamed S. Ebeida, and Scott A. Mitchell. A set of test problems and results in assessing method performance for calculating low probabilities of failure. In18th AIAA Non-Deterministic Approaches Conference, San Diego, CA, 2016. AIAA. doi: 10.2514/6.2016-0429. AIAA Paper 2016-0429

    work page doi:10.2514/6.2016-0429 2016

  58. [58]

    Rosenblatt

    M. Rosenblatt. Remarks on some nonparametric estimates of a density function.Ann. Math. Statist., 27(3):832–837, 1956

    work page 1956

  59. [59]

    Santner, Brian J

    Thomas J. Santner, Brian J. Williams, and William I. Notz.The Design and Analysis of Computer Experiments. Springer, 2003. doi: 10.1007/978-1-4757-3799-8

    work page doi:10.1007/978-1-4757-3799-8 2003

  60. [60]

    Sch¨ obi, B

    R. Sch¨ obi, B. Sudret, and S. Marelli. Rare event estimation using polynomial-chaos kriging.ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering, 3(2), 2017. doi: 10.1061/AJRUA6.0000870

    work page doi:10.1061/ajrua6.0000870 2017

  61. [61]

    D. W. Scott.Multivariate Density Estimation. Wiley, 2 edition, 2015

    work page 2015

  62. [62]

    B. W. Silverman.Density Estimation for Statistics and Data Analysis. Chapman & Hall, 1986

    work page 1986

  63. [63]

    The kernel method for univariate data

    Bernard W Silverman. The kernel method for univariate data. InDensity estimation for statistics and data analysis, pages 34–74. Springer, 1986

    work page 1986

  64. [64]

    Springer Science & Business Media, 2002

    Rajan Srinivasan.Importance sampling: Applications in communications and detection. Springer Science & Business Media, 2002

    work page 2002

  65. [65]

    G. R. Terrell and D. W. Scott. Variable kernel density estimation.Ann. Statist., 20(3): 1236–1265, 1992

    work page 1992

  66. [66]

    A. B. Tsybakov.Introduction to Nonparametric Estimation. Springer, 2009

    work page 2009

  67. [67]

    Veach and L

    E. Veach and L. J. Guibas. Optimally combining sampling techniques for monte carlo rendering. InProc. SIGGRAPH, pages 419–428, 1995

    work page 1995

  68. [68]

    M. P. Wand and M. C. Jones.Kernel Smoothing. Chapman & Hall, 1995

    work page 1995

  69. [69]

    Gaussian process surrogates for failure detection: A Bayesian experimental design approach.Journal of Computational Physics, 313:247–259, 2016

    Hongqiao Wang, Guang Lin, and Jinglai Li. Gaussian process surrogates for failure detection: A Bayesian experimental design approach.Journal of Computational Physics, 313:247–259, 2016. doi: 10.1016/j.jcp.2016.02.053

    work page doi:10.1016/j.jcp.2016.02.053 2016

  70. [70]

    Reliability analysis with stratified importance sampling based on adaptive kriging.Reliability Engineering & System Safety, 197:106852, 2020

    Sinan Xiao, Sergey Oladyshkin, and Wolfgang Nowak. Reliability analysis with stratified importance sampling based on adaptive kriging.Reliability Engineering & System Safety, 197:106852, 2020. doi: 10.1016/j.ress.2019.106852

    work page doi:10.1016/j.ress.2019.106852 2020

  71. [71]

    An active learning reliability method combining kriging and subset simulation.Reliability Engineering & System Safety, 188: 90–102, 2019

    Jing Zhang, Zhen Tong, Yiqun Li, and Jun Zhang. An active learning reliability method combining kriging and subset simulation.Reliability Engineering & System Safety, 188: 90–102, 2019. doi: 10.1016/j.ress.2019.03.001. 34

    work page doi:10.1016/j.ress.2019.03.001 2019

  72. [72]

    Response uncertainty and time-variant reliability analysis for hysteretic mdf structures.Earthquake engineering & structural dynamics, 28(10): 1187–1213, 1999

    YG Zhao, T Ono, and H Idota. Response uncertainty and time-variant reliability analysis for hysteretic mdf structures.Earthquake engineering & structural dynamics, 28(10): 1187–1213, 1999. 35

    work page 1999