Pith. sign in

REVIEW 3 major objections 4 minor 56 references

A novel stratified sampler with unbalanced refinement for network reliability assessment

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper proves that refining any stratum in a stratified network-reliability sampler cannot increase estimator variance relative to conditional Monte Carlo, and builds a practical sampler on that guarantee.

desk verdict Correct variance-monotonicity proof and a useful refinement construction, but the practical efficiency tables condition on an oracle for i*, which the GA often misses. read the letter →

arxiv 2506.01044 v1 pith:EOKZXUGT submitted 2025-06-01 stat.ME

Jianpeng Chan , Iason Papaioannou , Daniel Straub This is my paper · ORCID
classification stat.ME MSC 62D0562N0565C0590B25
keywords networkreliabilitystratifiedsamplingvariancereductionconditionalBernoullimodelsystemsignatureminimumfailurecardinalitygeneticalgorithmrareeventsimulation
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

Network reliability problems ask for the probability that a system of independent binary components fails, a number that is often tiny and expensive to estimate by simulation. This paper proposes a stratified sampler that splits the state space into progressively finer strata, each defined by a partition of the components into clusters and a count of failed components per cluster. The paper's central claim is that refining any stratum never increases the estimator's variance relative to conditional Monte Carlo when samples are allocated proportionally or optimally, so refinement can safely be pushed until computational cost becomes the limiting factor. The practical estimator also identifies the minimum number of failed components $i^*$ that can fail the network, discards all states below that threshold, and approximates the optimal per-stratum sample sizes under a coherence assumption. In a DC power-flow test and a water-supply connectivity test, the sampler reduces variance by orders of magnitude in the rare-event regime, with the noted exception of one threshold setting where approximation errors make it slightly less efficient than conditional Monte Carlo.

What carries the argument

The load-bearing object is the failure signature: a stratum is encoded by a partition of the $n$ components into clusters and a vector giving the number of failed components inside each cluster, and the conditional Bernoulli model supplies both the stratum probability (via the $R$ function) and exact conditional sampling (via the ID-checking sampler). The monotonicity proof in Appendix B reduces inequality (32a) to $(\lambda_{i2}a - \lambda_{i1}b)^2 \ge 0$ and inequality (32b) to $\big(\sqrt{a(\lambda_{i2}-b)} - \sqrt{b(\lambda_{i1}-a)}\big)^2 \ge 0$, where $a=\lambda_{i1}p_{F|i1}$ and $b=\lambda_{i2}p_{F|i2}$; this is what guarantees that refinement cannot hurt under proportional or optimal allocation. The second mechanism is the minimum-failure cardinality $i^*$: once it is known, states with fewer than $i^*$ failed components are provably safe and can be removed, turning the estimation problem into one with a much larger conditional failure probability $p_F^* = \Pr(F \mid I \ge i^*)$.

What would settle it

Take a small network with known failure probability by exhaustive enumeration, run the practical SSuR workflow with the genetic algorithm many times, and compare the GA-identified $i^*$ with the true value; if GA misidentifies $i^*$ in a non-negligible fraction of runs, the resulting relative bias should match the paper's reported values for equivalent problem parameters. Separately, for a physics-based metric, enumerate or sample all states that are supersets of observed failure states and test whether the coherence assumption holds; violations would show up as systematically misallocated samples and inflated $\alpha$.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes that stratified sampling for network reliability can be refined without risk: for any stratum $S_{i,j}$ with probability $\lambda_{i,j}$ and conditional failure probability $p_{F|i,j}$, splitting it into two sub-strata changes the variance ratios by amounts bounded by inequalities (32a) and (32b), and both bounds are non-positive. Equivalently, the proportional-allocation and optimal-allocation variance ratios $r^{\mathrm{SSuR,prop}}_{\mathrm{cMCS}}$ and $r^{\mathrm{SSuR,opt}}_{\mathrm{cMCS}}$ are non-increasing under refinement, with strict decrease unless the two sub-strata have identical conditional failure probabilities. The paper then combines this monotonicity with two practical devices: removal of all strata containing fewer than $i^*$ failed components, and a heuristic approximation of optimal sample sizes built from observed failure states and the assumption of a coherent network performance function. The resulting estimator, called SSuR, is demonstrated on an IEEE39 DC power-flow problem and a four-source water-supply network, where its variance relative to conditional Monte Carlo is reduced by factors ranging from about 1.2 to several thousand across the tested failure-probability regimes, with the practical version losing some of that gain to sample-size approximation and randomization.

Load-bearing premise

The estimator's unbiasedness rests on the genetic algorithm finding the true minimum number of failed components $i^*$ required for failure; if GA returns a larger value, failing states are silently excluded and the estimate is biased low, and the paper reports cases where this bias is close to half the true probability.

Editorial extensions

If this is right

  • Refinement can be applied iteratively without a variance-ratio penalty: each split either leaves the ratio unchanged or lowers it, so the only practical brake is the rising computational cost from one-sample strata.
  • When the probability mass below $i^*$ is large, removing redundant strata multiplies efficiency roughly by the ratio $p_F^*/p_F$, which in the examples reaches several orders of magnitude.
  • The optimal-allocation variance ratio can be much smaller than the proportional-allocation ratio, so the heuristic approximation of optimal sample sizes is the main source of practical loss; the paper quantifies that loss with $\alpha$ in Eq. (21).
  • For independent identical components, the stratified estimator is a direct estimator of the survival signature, since $p_{F|i} = 1 - \Phi_{n-i}$, so the refinement procedure doubles as a signature estimation method.

Reading between the lines

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

  • The non-increase guarantee is tied to proportional or optimal allocation; the paper's own figures show uniform allocation can push the variance ratio above one and even increase it under refinement, so practitioners should avoid uniform allocation with this scheme.
  • Because the genetic algorithm can miss the true $i^*$, the practical estimator is not always unbiased; for thr=50% with $p=5\times10^{-3}$ the reported relative bias is $-0.49$. A conservative alternative is to use a lower bound $i^*$ obtained by enumerating small failure counts, which restores unbiasedness at the cost of some efficiency.
  • The coherence assumption is testable: for a physics-based performance function, one could check whether every superset of an observed failure state is also a failure state, and if not, replace the sample-allocation heuristic with a non-coherent estimator.
  • The same refinement logic should extend to multi-state or dependent components if the conditional Bernoulli sampler is swapped for a suitable conditional sampler, since the monotonicity proof itself does not depend on binary inputs or independence.
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 proposes a stratified sampling estimator for network reliability, called SSuR, based on a partition of components into clusters and stratum refinement by numbers of failed components per cluster. The central theoretical result is that, under proportional or optimal allocation, refining any stratum never increases the variance ratio relative to (conditional) Monte Carlo; the proof is given in Appendix B. The practical estimator removes strata with fewer than i* failed components, approximates the optimal allocation using a coherency assumption and failure states observed by a genetic algorithm, and randomizes fractional sample sizes to integers. Efficiency is demonstrated on a DC power-flow example (IEEE39) and a water-supply connectivity example. The paper is transparent about several limitations, including that the physics-based numerical results assume i* is correctly identified and that the GA fails to identify i* in a substantial fraction of runs at one threshold.

Significance. If the claims hold, the paper makes a useful contribution: a clean, independent proof that stratified refinement cannot increase the variance ratio under proportional or optimal allocation, exact stratum-size computation via the conditional Bernoulli R-function, and a practical allocation heuristic. The proof of Eqs. (32a)-(32b) is mathematically sound, and the numerical studies use ground truth obtained by enumeration or MCS, not by the proposed method, so there is no circularity. However, the end-to-end practical efficiency is not demonstrated for the physics-based rare-event case at thr=50%, where the genetic algorithm identifies i* only 22% of the time, and the allocation approximation can be very inaccurate (alpha up to 55). The gap between the theoretical monotonicity result and the practical estimator's behavior is the main limitation.

major comments (3)
  1. [5.1.2, Tables 8-9] The headline practical efficiencies for the physics-based example are conditional on an oracle. Section 5.1.2 states that the stratified sampler results are computed assuming i* is correctly identified, while the only proposed procedure for finding i* for physics-based metrics is the GA, which in Table 8 has accuracy 0.22 at thr=50%. The text immediately following Table 8 reports relative biases of -0.49 at p=5e-3 and -0.29 at p=0.01 due to misidentification. A wrong i* truncates the wrong strata, so the estimator is biased and the MSE is not equal to the variance ratio reported in Table 9. Consequently, the claimed large efficiency gains for thr=50% are not end-to-end gains of the proposed workflow, and the concluding statement that the sampler 'outperforms clearly' crude and conditional MCS across all scenarios is not supported at that threshold. Please provide end-to-end results using the actual GA output (with MSE and bias reported), or adopt the unbiased lower-bound enumeration strategy mentioned in Section 3.3 and evaluate the resulting efficiency. For completeness, I note that for the other thresholds in Table 8 the GA accuracy is 1 and for the water-supply example i* is obtained exactly by max-flow, so the issue is localized but load-bearing for the physics-based rare-event claim.
  2. [5.1.2.2, Eq. (33), Tables 9-10] The practical allocation step can erase most of the theoretical gains. Table 10 reports alpha values up to 55 (thr=50%, p=1e-3) and 49 (thr=60%, p=1e-3), and Eq. (33) shows that the achievable efficiency is the optimal-allocation efficiency divided by (alpha+1). The measured relative efficiencies in Table 9 are consequently modest or below one at thr=60% (0.67-1.5 relative to conditional MCS). The paper acknowledges the thr=60% exception in the conclusions, but the discussion of 'large efficiency gains' is overstated for the practical estimator when alpha is large. The heuristic is also explicitly noted in Section 5.1.2.3 to estimate some positive conditional failure probabilities as zero, forcing a full sample into those strata; this is exactly the mechanism behind the large alpha values. Please discuss the practical implications of these alpha values and report the variability of the estimated relative efficiencies (e.g., confidence intervals) in addition to the point estimates from 10 runs.
  3. [2.3.3 and 4.5] The approximation of conditional failure probabilities for physics-based metrics relies on the coherency assumption, even though Section 4.5 acknowledges that physics-based performance functions are not always coherent. This is not a circularity issue, but it is a second fragile premise of the practical allocation: when non-coherent behavior is present, the estimated pF|i used in Eq. (9) can be systematically wrong, and the randomization rule in Eq. (23) then assigns a full sample to strata whose estimated probability is zero. The paper discloses the assumption but does not quantify its impact separately from the GA error. Please either quantify the effect of the coherency approximation in the examples, or state more precisely the class of performance functions for which the proposed practical allocation is expected to be reliable.
minor comments (4)
  1. [Eq. (23)] Equation (23) is rendered in a confusing way; the probabilities for the Bernoulli randomization of the sample size should be stated so that they clearly sum to one.
  2. [Section 5.2.2] The notation 'relEff opt,cMCS' and 'relEff prop,MCS' in the text of Section 5.2.2 is inconsistent with the estimator labels used elsewhere (bp(SSuR,aopt)_F and Table 12); please align the notation.
  3. [Section 4.4] The discussion of the number of refinement steps T correctly notes that the actual computational cost can exceed the initial sample size N, but the recommendation for choosing T under a fixed budget is only sketched; a more concrete algorithm or pseudocode would help reproducibility.
  4. [Section 5.1.2.3, Fig. 5] The caption of Figure 5 and the surrounding text describe relative efficiency and cost as functions of the refinement steps, but the cost curves are not clearly labeled in the text; please add a short legend description in the caption.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the variance-monotonicity claims are proven independently, and the numerical efficiency gap from GA misidentification of i* is a correctness limitation rather than a circular derivation.

full rationale

The paper's central mathematical claim, that the variance ratios r_SSuR,prop_cMCS and r_SSuR,opt_cMCS do not increase under stratum refinement, is derived from first principles in Appendix B (inequalities 32a and 32b), not imported from a self-citation. The text explicitly notes the property was 'also observed by Pettersson and Krumscheid [39]' but states 'we provide an independent and more detailed proof in Appendix B,' and the proof indeed reduces to completing squares, independent of the numerical examples or fitted parameters. The stratified estimator bp(SSuR)_F in Eq. (29) is defined directly from the total probability theorem, with stratum sizes computed from the conditional Bernoulli model; no estimated quantity is substituted into the definition of the quantity being predicted. The practical allocation approximation in Eq. (20) uses estimated conditional failure probabilities only to choose sample sizes; the estimator itself remains unbiased and its variance formula (Eq. 6) is the textbook stratified-sampling variance. The coherency-based approximation of pF|i,j affects efficiency but does not define the failure probability estimate. The acknowledged weakness, that Table 9 and Table 12 report efficiencies while 'assuming i* is correctly identified' while the GA achieves accuracy only 0.22 at thr = 50% (Table 8), is a real correctness and practical-efficiency concern, but it is not circularity: the reported variance ratios are computed under an explicit oracle assumption rather than being forced by construction from the method's own outputs. Ground truth in the examples is obtained by enumeration or MCS, not by the proposed estimator, so there is no fitted-input-called-prediction pattern. No load-bearing self-citation chain, uniqueness argument, or ansatz-smuggled-in-via-citation pattern is present. The paper's derivation chain is self-contained for its mathematical claims, and the numerical gaps are empirical correctness issues, not circular reasoning.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The derivation of the estimator and its variance properties introduces no fitted constants. The numerical demonstration, however, depends on hand-chosen GA hyperparameters, a fixed refinement budget T=5000, an initial sample size N=10000, and the coherency assumption for the allocation heuristic; these are the practical prices of the method.

free parameters (6)
  • GA population size npop = 500
    Selected from parameter study in Table 7; larger values improve i* identification accuracy but increase cost; the final examples use npop=500.
  • GA crossover fraction fxo = 0.8
    Chosen from parameter study; affects search behavior and accuracy.
  • GA tournament size ntrn = 2
    Fixed at 2; selected by the authors.
  • GA mutation rate pmt = 0.01
    Fixed; chosen by the authors.
  • Number of refinement steps T = 5000
    Set to 5000 in both examples; determines the granularity of strata and the computational cost; no automatic stopping criterion is used.
  • Initial sample size N = 10000
    Budget parameter for the stratified sampler in the numerical examples.
assumptions (5)
  • domain assumption Component states are independent and binary
    The method (Eq. 16, conditional Bernoulli model) is only valid for independent Bernoulli components; the paper states this and defers extensions.
  • domain assumption Network performance function is coherent
    Coherency is assumed in the allocation approximation (Section 2.3.3) and in the workflow (Section 4.5); the paper acknowledges it 'proceed[s] as if they were coherent' for physics-based functions.
  • ad hoc to paper The GA identifies the global minimum i*
    Sections 3.3 and 5.1.2.1; when the GA fails, the estimator is biased; the authors report accuracy rates below 1 for thr=50%.
  • standard math Fractional sample sizes are permitted for variance analysis
    The proofs of variance monotonicity and allocation formulas in Section 2 and Appendix B assume fractional Ni; integer randomization is handled later.
  • standard math R function and conditional Bernoulli sampling algorithms are correct
    The paper relies on published algorithms (Chen and Liu 1997, Chen et al. 1994) for computing Poisson-binomial probabilities and sampling; these are treated as established.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A novel stratified sampler with unbalanced refinement for network reliability assessment." pith.science (2026). https://pith.science/paper/EOKZXUGT

@misc{pith2026250601044,
  author       = {Pith},
  title        = {Pith review of: A novel stratified sampler with unbalanced refinement for network reliability assessment},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EOKZXUGT}},
  note         = {Machine review of arXiv:2506.01044}
}
abstract

We investigate stratified sampling in the context of network reliability assessment. We propose an unbalanced stratum refinement procedure, which operates on a partition of network components into clusters and the number of failed components within each cluster. The size of each refined stratum and the associated conditional failure probability, collectively termed failure signatures, can be calculated and estimated using the conditional Bernoulli model. The estimator is further improved by determining the minimum number of component failure $i^*$ to reach system failure and then by considering only strata with at least $i^*$ failed components. We propose a heuristic but practicable approximation of the optimal sample size for all strata, assuming a coherent network performance function. The efficiency of the proposed stratified sampler with unbalanced refinement (SSuR) is demonstrated through two network reliability problems.

Figures

Figures reproduced from arXiv: 2506.01044 by the authors.

Figure 1
Figure 1. A schematic plot of the refinement procedure. [PITH_FULL_IMAGE:figures/full_fig_p019_1.png] view at source ↗
Figure 2
Figure 2. Workflow of the proposed stratified sampler. [PITH_FULL_IMAGE:figures/full_fig_p020_2.png] view at source ↗
Figure 3
Figure 3. Variance ratios of the stratified sampler, [PITH_FULL_IMAGE:figures/full_fig_p026_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Variance ratios of the stratified sampler, [PITH_FULL_IMAGE:figures/full_fig_p026_4.png]
Figure 5
Figure 5. Figure 5: Influence of the number of refinement steps on relEff [PITH_FULL_IMAGE:figures/full_fig_p030_5.png]
Figure 6
Figure 6. Figure 6: Topology of the water supply system in Example 5.2. [PITH_FULL_IMAGE:figures/full_fig_p032_6.png]
Figure 7
Figure 7. Figure 7: Variance ratios of the stratified sampler, [PITH_FULL_IMAGE:figures/full_fig_p033_7.png]
Figure 8
Figure 8. Figure 8: Variance ratios of the stratified sampler, [PITH_FULL_IMAGE:figures/full_fig_p034_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 56 canonical work pages

  1. [1]

    Li and W

    J. Li and W. Liu, Lifeline Engineering Systems: Network Reliability Analysis and Aseismic Design . Springer Nature, 2021

    work page 2021

  2. [2]

    Billinton and W

    R. Billinton and W. Li, Reliability assessment of electric power systems using Monte Carlo methods . Springer Science & Business Media, 1994

    work page 1994

  3. [3]

    Computing network reliability in time polynomial in the number of cuts,

    J. S. Provan and M. O. Ball, “Computing network reliability in time polynomial in the number of cuts,” Operations Research, vol. 32, no. 3, pp. 516–526, 1984

    work page 1984

  4. [4]

    Computational complexity of network reliability analysis: An overview,

    M. O. Ball, “Computational complexity of network reliability analysis: An overview,” IEEE Transactions on Reliability, vol. 35, no. 3, pp. 230– 239, 1986

    work page 1986

  5. [5]

    Reliability evaluation of a limited- flow network in terms of minimal cutsets,

    C.-C. Jane, J.-S. Lin, and J. Yuan, “Reliability evaluation of a limited- flow network in terms of minimal cutsets,” IEEE Transactions on Reli- ability, vol. 42, no. 3, pp. 354–361, 1993

    work page 1993

  6. [6]

    An efficient method for reliability evaluation of multistate networks given all minimal path vectors,

    M. J. Zuo, Z. Tian, and H.-Z. Huang, “An efficient method for reliability evaluation of multistate networks given all minimal path vectors,” IIE Transactions, vol. 39, no. 8, pp. 811–817, 2007

    work page 2007

  7. [7]

    Network reliability: Heading out on the highway,

    J. I. Brown, C. J. Colbourn, D. Cox, C. Graves, and L. Mol, “Network reliability: Heading out on the highway,” Networks, vol. 77, no. 1, pp. 146–160, 2021

    work page 2021

  8. [8]

    Computational investigations of all-terminal network reliability via BDDs,

    H. Imai, K. Sekine, and K. Imai, “Computational investigations of all-terminal network reliability via BDDs,” IEICE Transactions on Fundamentals of Electronics, Communications and Computer Sciences , vol. 82, no. 5, pp. 714–721, 1999

    work page 1999

Show all 56 references
  1. [9]

    K-terminal network reliability measures with binary decision diagrams,

    G. Hardy, C. Lucet, and N. Limnios, “K-terminal network reliability measures with binary decision diagrams,” IEEE Transactions on Relia- bility, vol. 56, no. 3, pp. 506–515, 2007

    2007

  2. [10]

    Generalised importance mea- sures for multi-state elements based on performance level restrictions,

    G. Levitin, L. Podofillini, and E. Zio, “Generalised importance mea- sures for multi-state elements based on performance level restrictions,” Reliability Engineering & System Safety , vol. 82, no. 3, pp. 287–298, 2003. 40

    2003

  3. [11]

    A multi-state model for the reliability assessment of a distributed generation system via universal generating function,

    Y.-F. Li and E. Zio, “A multi-state model for the reliability assessment of a distributed generation system via universal generating function,” Reliability Engineering & System Safety , vol. 106, pp. 28–36, 2012

    2012

  4. [12]

    System reliability and sensitivity under statis- tical dependence by matrix-based system reliability method,

    J. Song and W.-H. Kang, “System reliability and sensitivity under statis- tical dependence by matrix-based system reliability method,” Structural Safety, vol. 31, no. 2, pp. 148–156, 2009

    2009

  5. [13]

    A recursive decomposition algorithm for network seis- mic reliability evaluation,

    J. Li and J. He, “A recursive decomposition algorithm for network seis- mic reliability evaluation,” Earthquake Engineering & Structural Dy- namics, vol. 31, no. 8, pp. 1525–1539, 2002

    2002

  6. [14]

    Efficient risk assessment of lifeline networks under spatially correlated ground motions using selective recursive de- composition algorithm,

    H.-W. Lim and J. Song, “Efficient risk assessment of lifeline networks under spatially correlated ground motions using selective recursive de- composition algorithm,” Earthquake Engineering & Structural Dynam- ics, vol. 41, no. 13, pp. 1861–1882, 2012

    2012

  7. [15]

    Decompo- sition algorithms for system reliability estimation with applications to interdependent lifeline networks,

    R. Paredes, L. Due˜ nas-Osorio, and I. Hernandez-Fajardo, “Decompo- sition algorithms for system reliability estimation with applications to interdependent lifeline networks,” Earthquake Engineering & Structural Dynamics, vol. 47, no. 13, pp. 2581–2600, 2018

    2018

  8. [16]

    Branch-and-bound algorithm for efficient reliability analysis of general coherent systems,

    J.-E. Byun, H. Ryu, and D. Straub, “Branch-and-bound algorithm for efficient reliability analysis of general coherent systems,” arXiv preprint arXiv:2410.22363, 2024

    2024 arXiv

  9. [17]

    Reliability analysis of discrete multi-state sys- tems by means of subset simulation,

    E. Zio and N. Pedroni, “Reliability analysis of discrete multi-state sys- tems by means of subset simulation,” in Proceedings of the 17th ESREL Conference, Valencia, Spain, 2008, pp. 22–25

    2008

  10. [18]

    General network reliability problem and its efficient solution by subset simulation,

    K. M. Zuev, S. Wu, and J. L. Beck, “General network reliability problem and its efficient solution by subset simulation,”Probabilistic Engineering Mechanics, vol. 40, pp. 25–35, 2015

    2015

  11. [19]

    A stochastic framework for reliability and sensitivity analysis of large scale water distribution networks,

    H. A. Jensen and D. J. Jerez, “A stochastic framework for reliability and sensitivity analysis of large scale water distribution networks,” Re- liability Engineering & System Safety , vol. 176, pp. 80–92, 2018

    2018

  12. [20]

    An adaptive subset simulation algorithm for system reliability analysis with discontinuous limit states,

    J. Chan, I. Papaioannou, and D. Straub, “An adaptive subset simulation algorithm for system reliability analysis with discontinuous limit states,” Reliability Engineering & System Safety , vol. 225, p. 108607, 2022. 41

    2022

  13. [21]

    The cross-entropy method for network reliability estimation,

    K.-P. Hui, N. Bean, M. Kraetzl, and D. P. Kroese, “The cross-entropy method for network reliability estimation,” Annals of Operations Re- search, vol. 134, no. 1, pp. 101–118, 2005

    2005

  14. [22]

    Bayesian improved cross en- tropy method for network reliability assessment,

    J. Chan, I. Papaioannou, and D. Straub, “Bayesian improved cross en- tropy method for network reliability assessment,” Structural Safety, vol. 103, p. 102344, 2023

    2023

  15. [23]

    Bayesian improved cross entropy method with categorical mix- ture models,

    ——, “Bayesian improved cross entropy method with categorical mix- ture models,” Reliability Engineering & System Safety , vol. 252, p. 110432, 2024

    2024

  16. [24]

    Network reliability analysis: Part 1,

    R. Van Slyke and H. Frank, “Network reliability analysis: Part 1,” Net- works, vol. 1, no. 3, pp. 279–290, 1971

    1971

  17. [25]

    Adaptive monte carlo methods for estimating rare events in power grids,

    J. Chan, R. Paredes, I. Papaioannou, L. Duenas-Osorio, and D. Straub, “Adaptive monte carlo methods for estimating rare events in power grids,” ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A: Civil Engineering , vol. 11, no. 1, p. 04024082, 2025

    2025

  18. [26]

    Counting-based reliability estimation for power-transmission grids,

    L. Duenas-Osorio, K. Meel, R. Paredes, and M. Vardi, “Counting-based reliability estimation for power-transmission grids,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 31, no. 1, San Francisco, California, USA, 2017

    2017

  19. [27]

    Estimation of network reliability using graph evolution models,

    T. Elperin, I. Gertsbakh, and M. Lomonosov, “Estimation of network reliability using graph evolution models,” IEEE Transactions on Relia- bility, vol. 40, no. 5, pp. 572–581, 1991

    1991

  20. [28]

    Reliability estimation for stochastic flow networks with dependent arcs,

    H. Cancela, L. Murray, and G. Rubino, “Reliability estimation for stochastic flow networks with dependent arcs,” IEEE Transactions on Reliability, pp. 622–636, 2022

    2022

  21. [29]

    A recursive variance-reduction algo- rithm for estimating communication-network reliability,

    H. Cancela and M. El Khadiri, “A recursive variance-reduction algo- rithm for estimating communication-network reliability,” IEEE Trans- actions on Reliability , vol. 44, no. 4, pp. 595–602, 1995

    1995

  22. [30]

    Adaptive network reliability analysis: Methodology and applications to power grid,

    N. L. Dehghani, S. Zamanian, and A. Shafieezadeh, “Adaptive network reliability analysis: Methodology and applications to power grid,” Reli- ability Engineering & System Safety , vol. 216, p. 107973, 2021. 42

    2021

  23. [31]

    Sampling and active learning methods for network reliability estimation using K- terminal spanning tree,

    C. Ding, P. Wei, Y. Shi, J. Liu, M. Broggi, and M. Beer, “Sampling and active learning methods for network reliability estimation using K- terminal spanning tree,” Reliability Engineering & System Safety , vol. 250, p. 110309, 2024

    2024

  24. [32]

    Survival signature for reliability quantification of large systems and networks,

    F. P. Coolen and T. Coolen-Maturi, “Survival signature for reliability quantification of large systems and networks,” in System Dependability - Theory and Applications , W. Zamojski, J. Mazurkiewicz, J. Sugier, T. Walkowiak, and J. Kacprzyk, Eds. Springer Nature Switzerland, 2...

    2024

  25. [33]

    Entropy-driven Monte Carlo simulation method for approximating the survival signature of complex infrastructures,

    F. Di Maio, C. Pettorossi, and E. Zio, “Entropy-driven Monte Carlo simulation method for approximating the survival signature of complex infrastructures,” Reliability Engineering & System Safety , vol. 231, p. 108982, 2023

    2023

  26. [34]

    Refinement strategies for stratified sampling methods,

    C. Tong, “Refinement strategies for stratified sampling methods,” Reli- ability Engineering & System Safety , vol. 91, no. 10-11, pp. 1257–1265, 2006

    2006

  27. [35]

    Refined strat- ified sampling for efficient Monte Carlo based uncertainty quantifica- tion,

    M. D. Shields, K. Teferra, A. Hapij, and R. P. Daddazio, “Refined strat- ified sampling for efficient Monte Carlo based uncertainty quantifica- tion,” Reliability Engineering & System Safety , vol. 142, pp. 310–325, 2015

    2015

  28. [36]

    W. G. Cochran, Sampling Techniques. John Wiley & Sons, 1977

    1977

  29. [37]

    Adaptive stratified sampling for structural re- liability analysis,

    C. Song and R. Kawai, “Adaptive stratified sampling for structural re- liability analysis,” Structural Safety, vol. 101, p. 102292, 2023

    2023

  30. [38]

    On adaptive stratifi- cation,

    P. Etor´ e, G. Fort, B. Jourdain, and E. Moulines, “On adaptive stratifi- cation,” Annals of operations research, vol. 189, pp. 127–154, 2011

    2011

  31. [39]

    Adaptive stratified sampling for non- smooth problems,

    P. Pettersson and S. Krumscheid, “Adaptive stratified sampling for non- smooth problems,” International Journal for Uncertainty Quantifica- tion, vol. 12, no. 6, pp. 71–99, 2022

    2022

  32. [40]

    Monte Carlo estimation of the maximal flow distribu- tion with discrete stochastic arc capacity levels,

    G. S. Fishman, “Monte Carlo estimation of the maximal flow distribu- tion with discrete stochastic arc capacity levels,” Naval Research Logis- tics (NRL) , vol. 36, no. 6, pp. 829–849, 1989. 43

    1989

  33. [41]

    Network reliability es- timation using the tree cut and merge algorithm with importance sam- pling,

    K.-P. Hui, N. Bean, M. Kraetzl, and D. Kroese, “Network reliability es- timation using the tree cut and merge algorithm with importance sam- pling,” in Proceedings of the 4th International Workshop on Design of Reliable Communication Networks . Banff, Alberta, Canada: IEEE, 20...

    2003

  34. [42]

    Lecture notes in Advanced Stochastic Finite Element Methods,

    I. Papaioannou, “Lecture notes in Advanced Stochastic Finite Element Methods,” Munich, Germany, 2021

    2021

  35. [43]

    Statistical applications of the Poisson- binomial and conditional Bernoulli distributions,

    S. X. Chen and J. S. Liu, “Statistical applications of the Poisson- binomial and conditional Bernoulli distributions,” Statistica Sinica , vol. 7, no. 4, pp. 875–892, 1997

    1997

  36. [44]

    Weighted finite population sampling to maximize entropy,

    X.-H. Chen, A. P. Dempster, and J. S. Liu, “Weighted finite population sampling to maximize entropy,” Biometrika, vol. 81, no. 3, pp. 457–469, 1994

    1994

  37. [45]

    Likelihood calculations for matched case-control studies and survival studies with tied death times,

    M. H. Gail, J. H. Lubin, and L. V. Rubinstein, “Likelihood calculations for matched case-control studies and survival studies with tied death times,” Biometrika, vol. 68, no. 3, pp. 703–707, 1981

    1981

  38. [46]

    A new technique in minimal path and cutset evaluation,

    G. B. Jasmon and O. S. Kai, “A new technique in minimal path and cutset evaluation,” IEEE Transactions on Reliability , vol. R-34, no. 2, pp. 136–143, 1985

    1985

  39. [47]

    Maximal flow through a network,

    L. R. Ford and D. R. Fulkerson, “Maximal flow through a network,” Canadian Journal of Mathematics , vol. 8, pp. 399–404, 1956

    1956

  40. [48]

    A simple min-cut algorithm,

    M. Stoer and F. Wagner, “A simple min-cut algorithm,” Journal of the ACM (JACM), vol. 44, no. 4, pp. 585–591, 1997

    1997

  41. [49]

    Non-convex mixed-integer nonlinear pro- gramming: A survey,

    S. Burer and A. N. Letchford, “Non-convex mixed-integer nonlinear pro- gramming: A survey,” Surveys in Operations Research and Management Science, vol. 17, no. 2, pp. 97–106, 2012

    2012

  42. [50]

    Mitchell, An Introduction to Genetic Algorithms

    M. Mitchell, An Introduction to Genetic Algorithms . MIT press, 1998

    1998

  43. [51]

    A fast algorithm for searching all multi-state minimal cuts,

    W.-C. Yeh, “A fast algorithm for searching all multi-state minimal cuts,” IEEE Transactions on Reliability , vol. 57, no. 4, pp. 581–588, 2008. 44

    2008

  44. [52]

    Efficiency improvement and variance reduction,

    P. L’Ecuyer, “Efficiency improvement and variance reduction,” in Pro- ceedings of Winter Simulation Conference. San Diego, CA, USA: IEEE, 1994, pp. 122–132

    1994

  45. [53]

    J. J. Grainger, Power System Analysis . McGraw-Hill, 1999

    1999

  46. [54]

    Model for cascading failures in complex networks,

    P. Crucitti, V. Latora, and M. Marchiori, “Model for cascading failures in complex networks,” Physical Review E, vol. 69, no. 4, p. 045104, 2004

    2004

  47. [55]

    The seismic response and dynamic function reliability analysis of underground water supply networks (in Chinese),

    H. Miao, “The seismic response and dynamic function reliability analysis of underground water supply networks (in Chinese),” PhD thesis, Tongji University, 2018

    2018

  48. [56]

    A preliminary research on the damage prediction of the buried line (in Chinese),

    D. Wang, “A preliminary research on the damage prediction of the buried line (in Chinese),” Journal of Zhengzhou Institute of Technol- ogy, vol. 12, no. 1, pp. 65–68, 1991. 45

    1991

Pith tools

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