pith. sign in

arxiv: 2605.04751 · v1 · submitted 2026-05-06 · 📡 eess.SY · cs.SY

Sequential Monte Carlo for Resilient Networks: Assessment, Mitigation, and Generative Modeling

Onel L. A. L\'opez , Amirhossein Azarbahram This is my paper

Pith reviewed 2026-05-08 17:10 UTC · model grok-4.3

classification 📡 eess.SY cs.SY
keywords sequential monte carloresilient networksrare-event simulationwireless networksgenerative modelingdigital twinsnetwork resilience
0
0 comments X

The pith

Sequential Monte Carlo with fixed-level splitting estimates rare non-recovery probabilities in wireless networks more efficiently than standard Monte Carlo.

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

This paper develops a sequential Monte Carlo simulation framework for assessing resilience in future wireless networks by formulating failures as path-dependent rare events from staged degradation and delayed recovery. These events are decomposed into semantically interpretable levels using a reaction coordinate. The framework applies fixed-level splitting with budget-aware population control to estimate rare non-recovery probabilities efficiently and reuses checkpoints as near-critical states for policy evaluation. It further extends the approach to learned simulators by treating generative sequence models as restartable surrogates in data-driven digital twins. In a delay-critical wireless network example, the method shows substantial improvement over standard Monte Carlo for both physical and learned simulators.

Core claim

Resilience failures are formulated as path-dependent rare events arising from staged degradation and delayed recovery and decomposed into semantically interpretable levels defined by a reaction coordinate. A fixed-level splitting approach with budget-aware population control then enables efficient estimation of rare non-recovery probabilities. SMC checkpoints can be reused as representative near-critical states for policy evaluation and simulation-based selection. The methodology extends to learned stochastic simulation by using generative sequence models as restartable surrogates within data-driven digital twins, as demonstrated in a delay-critical wireless network use case where SMC yields

What carries the argument

Fixed-level splitting approach with budget-aware population control within sequential Monte Carlo, which decomposes path-dependent rare events via a reaction coordinate to enable efficient probability estimation and checkpoint reuse.

If this is right

  • SMC checkpoints serve as representative near-critical states for policy evaluation and simulation-based selection.
  • The framework applies equally to physical simulators and to learned stochastic simulators via generative sequence models.
  • Efficient rare-event estimation supports quantitative assessment and mitigation of cascading disruptions in wireless networks.

Where Pith is reading between the lines

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

  • The reaction coordinate decomposition could transfer to resilience modeling in other infrastructures such as power grids or transportation systems.
  • Dynamic adjustment of splitting levels using online observations might further reduce the computational budget required for large-scale networks.
  • Integration with real-time digital twins could enable proactive resilience policies triggered by near-critical state detection.

Load-bearing premise

Resilience failures can be represented as sequences of staged degradation and delayed recovery that decompose into distinct levels via a reaction coordinate.

What would settle it

Running both the SMC framework and standard Monte Carlo on an identical delay-critical wireless network simulation with a pre-computed or known rare non-recovery probability and checking whether SMC produces lower-variance estimates or faster convergence at small probability values.

Figures

Figures reproduced from arXiv: 2605.04751 by Amirhossein Azarbahram, Onel L. A. L\'opez.

Figure 1
Figure 1. Figure 1: Illustration of system dynamics and trajectory view at source ↗
Figure 2
Figure 2. Figure 2: Illustration of the online mitigation/control pro view at source ↗
Figure 3
Figure 3. Figure 3: Sequence generation modeling choices for a view at source ↗
Figure 4
Figure 4. Figure 4: System dynamics of the use case. Here, ν(u) > 0 is the recovery rate and φ(u) > 1 controls the nonlinearity of the recovery dynamics, both controlled by a recovery acceleration action u. The recovery rate models adaptive system response that increases when the system operates below capacity, while the exponential term captures nonlinear amplification of stress effects, consistent with models of cascading f… view at source ↗
Figure 6
Figure 6. Figure 6: Estimated non-recovery probability Pr(ξ) versus the standard deviation of the latent stress σF with the proposed SMC-assisted reconfiguration framework and policy sets U of different dimensions. We set ρ ′ = 1/2, κ = 1/2, and N′ = 25. For the case of |U| = 5, we also plot the relative selection frequency of each policy for σF ∈ {0.45, 0.575, 0.8}. selection rule introduced in Section III-B, such that (25) … view at source ↗
Figure 5
Figure 5. Figure 5: Estimated non-recovery probability Pr(ξ) versus the service degradation threshold δ for Λ ∈ {0.6, 0.7} ms and using MC and SMC simulation approaches. C. Service Reconfiguration Now, we showcase SMC for fault management as discussed in Section III-B. For simplicity, we consider service reconfiguration actions u that are activated only after the system enters a critically degraded operating regime, correspon… view at source ↗
Figure 7
Figure 7. Figure 7: Comparison between physical-model and DDPM view at source ↗
Figure 8
Figure 8. Figure 8: compares the estimated non-recovery probability obtained from the physical simulator and the DDPM surrogate using MC and SMC sampling. The DDPM es￾timates closely follow the physical-model baselines across the load range. The benefit of SMC is most visible in the low-probability region, where DDPM-MC is limited by the finite number of generated trajectories, while view at source ↗
read the original abstract

Resilience is becoming crucial for future wireless networks, which must withstand, adapt to, and recover from rare but potentially cascading disruptions. This paper develops a sequential Monte Carlo (SMC) simulation framework for such systems, in which resilience failures are formulated as path-dependent rare events arising from staged degradation and delayed recovery, and are decomposed into semantically interpretable levels defined by a reaction coordinate. Building on this structure, we present a fixed-level splitting approach with budget-aware population control, enabling efficient estimation of rare non-recovery probabilities. We discuss the potential reuse of SMC checkpoints as representative near-critical states for policy evaluation and simulation-based selection. We further extend the methodology to learned stochastic simulation by using generative sequence models as restartable surrogates within data-driven digital twins. We showcase the framework in a delay-critical wireless network use case, where SMC substantially improves over standard Monte Carlo in rare-event regimes with both physical and learned simulators.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The paper develops a sequential Monte Carlo (SMC) framework for resilience assessment in wireless networks by formulating failures as path-dependent rare events from staged degradation and delayed recovery, decomposed via a reaction coordinate into interpretable levels. It introduces fixed-level splitting with budget-aware population control for efficient rare-event probability estimation, discusses reuse of SMC checkpoints for policy evaluation, and extends the approach to learned simulators by treating generative sequence models as restartable surrogates in data-driven digital twins. The framework is demonstrated on a delay-critical wireless network, claiming substantial improvements over standard Monte Carlo for both physical and learned simulators.

Significance. If the central claims hold, the work offers a principled simulation-based approach to rare-event resilience analysis that could improve efficiency in network design and digital-twin applications. The extension to generative models as surrogates, if validated, would be a notable contribution for data-driven settings where physical simulators are costly.

major comments (1)
  1. [Abstract and generative-modeling extension] The extension to learned simulators (abstract and generative-modeling section) treats generative sequence models as faithful restartable surrogates inside the SMC loop, but provides no explicit validation that the surrogates preserve reaction-coordinate level crossings or non-recovery probabilities on staged degradation paths. Since these models are trained on nominal trajectories, distortion of rare-event tails remains an untested assumption that directly undermines the reported improvement for learned simulators.
minor comments (2)
  1. [Methodology] Clarify the precise definition and selection procedure for the reaction coordinate levels, including any sensitivity analysis with respect to the free parameters listed in the axiom ledger.
  2. [Numerical results] Add quantitative comparison metrics (e.g., variance reduction factors or effective sample sizes) between SMC and standard Monte Carlo across multiple rare-event regimes to substantiate the 'substantial improvement' claim.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We address the major comment point by point below and outline the revisions we will make.

read point-by-point responses

  1. Referee: [Abstract and generative-modeling extension] The extension to learned simulators (abstract and generative-modeling section) treats generative sequence models as faithful restartable surrogates inside the SMC loop, but provides no explicit validation that the surrogates preserve reaction-coordinate level crossings or non-recovery probabilities on staged degradation paths. Since these models are trained on nominal trajectories, distortion of rare-event tails remains an untested assumption that directly undermines the reported improvement for learned simulators.

    Authors: We agree with the referee that the manuscript does not include explicit numerical validation demonstrating that the generative sequence models preserve the reaction-coordinate level crossings or the non-recovery probabilities along staged degradation paths. The current presentation treats the learned simulators as restartable surrogates in a conceptual extension, with the reported improvements for the learned-simulator case relying on the assumption that the models trained on nominal trajectories remain sufficiently accurate in the rare-event regime. This is indeed an untested assumption that warrants direct verification. In the revised manuscript we will add a dedicated subsection in the generative-modeling section that compares SMC estimates (including level-crossing frequencies and final non-recovery probabilities) obtained from the physical simulator against those obtained from the learned surrogate on identical rare-event scenarios. We will report quantitative fidelity metrics for the reaction coordinate and discuss any observed tail distortion. This addition will directly address the concern and strengthen the claims for the data-driven digital-twin setting. revision: yes

Circularity Check

0 steps flagged

No circularity in SMC rare-event framework derivation

full rationale

The paper applies standard sequential Monte Carlo with fixed-level splitting and population control to path-dependent rare events defined via an external reaction coordinate. The extension to generative sequence models treats them as restartable surrogates trained on separate data, without any self-definition, fitted-input-as-prediction, or load-bearing self-citation that reduces the central claims to tautology. All load-bearing steps (formulation of non-recovery probabilities, budget-aware control, surrogate reuse) remain externally verifiable against physical simulators and independent benchmarks. No equations or steps in the provided text collapse by construction to their inputs.

Axiom & Free-Parameter Ledger

2 free parameters · 1 axioms · 0 invented entities

Based on abstract only; the framework relies on standard SMC assumptions plus domain modeling of network degradation and recovery. No new entities are introduced.

free parameters (2)
  • reaction coordinate levels
    Thresholds for decomposing rare events into levels, chosen to enable splitting.
  • population control budget parameters
    Controls for managing simulation particles at each stage in budget-aware splitting.
axioms (1)
  • domain assumption Resilience failures can be modeled as path-dependent rare events decomposable via a reaction coordinate.
    Foundational modeling choice stated in the abstract for applying SMC.

pith-pipeline@v0.9.0 · 5465 in / 1338 out tokens · 33007 ms · 2026-05-08T17:10:23.637192+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

36 extracted references · 36 canonical work pages · 2 internal anchors

  1. [1]

    Rak and D

    J. Rak and D. Hutchison, Guide to disaster-resilient commu- nication networks. Springer Nature, 2020

    work page 2020

  2. [2]

    Resilient radio access net- works: Ai and the unknown unknowns,

    B. Matthiesen et al., “Resilient radio access networks: AI and the unknown unknowns,” arXiv preprint arXiv:2510.21587, 2025

    work page arXiv 2025

  3. [3]

    Alves, et al., 6G Resilience White Paper [White paper]

    H. Alves, et al., 6G Resilience White Paper [White paper]. 6G Research Visions, nr. 15, Oulu, Finland: University of Oulu, Nov. 2025

    work page 2025

  4. [4]

    Generative AI empowered network digital twins: Architecture, technologies, and applications,

    T. Li et al., “Generative AI empowered network digital twins: Architecture, technologies, and applications,” ACM Comput. Surv., vol. 57, Feb. 2025

    work page 2025

  5. [5]

    Modelling for digital twins-potential role of surrogate models,

    A. Bárkányi et al., “Modelling for digital twins-potential role of surrogate models,” Processes, vol. 9, no. 3, 2021

    work page 2021

  6. [6]

    Auto-Encoding Variational Bayes

    D. Kingma and M. Welling, “Auto-encoding variational bayes,” arXiv preprint arXiv:1312.6114, 2013

    work page internal anchor Pith review arXiv 2013

  7. [7]

    Normalizing flows for probabilistic modeling and inference,

    G. Papamakarios et al., “Normalizing flows for probabilistic modeling and inference,” J. Mach. Learn. Res., vol. 22, no. 57, pp. 1–64, 2021

    work page 2021

  8. [8]

    Time-series generative adversarial networks,

    J. Yoon et al., “Time-series generative adversarial networks,” in NeurIPS, vol. 32, Curran Associates, Inc., 2019

    work page 2019

  9. [9]

    Denoising diffusion probabilistic models,

    J. Ho et al., “Denoising diffusion probabilistic models,” in NeurIPS, vol. 33, pp. 6840–6851, Curran Associates, Inc., 2020

    work page 2020

  10. [10]

    Rubino et al., Rare event simulation using Monte Carlo methods, vol

    G. Rubino et al., Rare event simulation using Monte Carlo methods, vol. 73. Wiley Online Library, 2009

    work page 2009

  11. [11]

    Statistical tools and methodologies for ultrareliable low-latency communication—a tutorial,

    O. López, et al., “Statistical tools and methodologies for ultrareliable low-latency communication—a tutorial,” Proc. IEEE, vol. 111, no. 11, pp. 1502–1543, 2023

    work page 2023

  12. [12]

    Sequential Monte Carlo for network resilience as- sessment and control,

    O. López, “Sequential Monte Carlo for network resilience as- sessment and control,” arXiv preprint arXiv:2604.00540, 2026

    work page arXiv 2026

  13. [13]

    Sequential Monte Carlo samplers,

    P. Moral et al., “Sequential Monte Carlo samplers,” Journal of the Royal Statistical Society Series B: Statistical Methodology, vol. 68, no. 3, pp. 411–436, 2006

    work page 2006

  14. [14]

    Smith, Sequential Monte Carlo methods in practice

    A. Smith, Sequential Monte Carlo methods in practice. Springer Science & Business Media, 2013

    work page 2013

  15. [15]

    RESTART: a method for accel- erating rare event simulations,

    M. Villén-Altamirano et al., “RESTART: a method for accel- erating rare event simulations,” Queueing, Performance and Control in ATM (ITC-13), pp. 71–76, 1991

    work page 1991

  16. [16]

    The splitting method in rare event simulation,

    M. Garvels, “The splitting method in rare event simulation,” 2000

    work page 2000

  17. [17]

    On the importance function in splitting simulation,

    M. Garvels et al., “On the importance function in splitting simulation,” Eur. Trans. Telecommun., vol. 13, no. 4, pp. 363– 371, 2002

    work page 2002

  18. [18]

    Bayesian subset simulation,

    J. Bect et al., “Bayesian subset simulation,” SIAM/ASA J. Uncertain. Quantif., vol. 5, no. 1, pp. 762–786, 2017

    work page 2017

  19. [19]

    Adaptive multilevel splitting for rare event analysis,

    F. Cérou and A. Guyader, “Adaptive multilevel splitting for rare event analysis,” Stoch. Anal. Appl., vol. 25, no. 2, pp. 417– 443, 2007

    work page 2007

  20. [20]

    An efficient statistical model checker for nondeterminism and rare events,

    C. Budde et al., “An efficient statistical model checker for nondeterminism and rare events,” Int. J. Softw. Tools Technol. Transfer, vol. 22, no. 6, pp. 759–780, 2020

    work page 2020

  21. [21]

    Fast simulation of ultra-reliable coded com- munication system via adaptive shaping of noise histogram,

    Y. Yu and D. Lin, “Fast simulation of ultra-reliable coded com- munication system via adaptive shaping of noise histogram,” in IEEE VTC-Spring), pp. 1–5, 2020

    work page 2020

  22. [22]

    Adaptive importance sampling and quasi- Monte Carlo methods for 6G URLLC systems,

    X. Ke et al., “Adaptive importance sampling and quasi- Monte Carlo methods for 6G URLLC systems,” in IEEE ICC, pp. 5272–5278, 2023

    work page 2023

  23. [23]

    Fast simulation of channel-coded MIMO-OFDM transmission in multipath fading at low error rates,

    Y. Yu et al., “Fast simulation of channel-coded MIMO-OFDM transmission in multipath fading at low error rates,” IEEE Trans. Commun., vol. 73, no. 8, pp. 6351–6365, 2025

    work page 2025

  24. [24]

    A flow-based generative model for rare-event simulation,

    L. Gibson et al., “A flow-based generative model for rare-event simulation,” arXiv preprint arXiv:2305.07863, 2023

    work page arXiv 2023

  25. [25]

    Rare event probability learning by normalizing flows,

    Z. Gao et al., “Rare event probability learning by normalizing flows,” arXiv preprint arXiv:2310.19167, 2023

    work page arXiv 2023

  26. [26]

    REIN: reliability estimation via importance sampling with normalizing flows,

    A. Dasgupta and E. Johnson, “REIN: reliability estimation via importance sampling with normalizing flows,” Reliab. Eng. Syst. Saf., vol. 242, p. 109729, 2024

    work page 2024

  27. [27]

    FIDE: frequency-inflated conditional diffusion model for extreme-aware time series generation,

    A. Galib et al., “FIDE: frequency-inflated conditional diffusion model for extreme-aware time series generation,” in NeurIPS, vol. 37, pp. 114434–114457, Curran Associates, Inc., 2024

    work page 2024

  28. [28]

    Can diffusion models capture extreme event statistics?,

    S. Stamatelopoulos and T. Sapsis, “Can diffusion models capture extreme event statistics?,” Comput. Methods Appl. Mech. Eng., vol. 435, p. 117589, 2025

    work page 2025

  29. [29]

    Heavy-tailed diffusion models,

    K. Pandey et al., “Heavy-tailed diffusion models,” in ICLR, 2024

    work page 2024

  30. [30]

    Efficient rare event sampling with unsuper- vised normalizing flows,

    S. Asghar et al., “Efficient rare event sampling with unsuper- vised normalizing flows,” Nat. Mach. Intell., vol. 6, pp. 1370– 1381, Nov 2024

    work page 2024

  31. [31]

    Analysis of adaptive multilevel splitting algorithms in an idealized case,

    C.-E. Bréhier et al., “Analysis of adaptive multilevel splitting algorithms in an idealized case,” ESAIM: P&S, vol. 19, pp. 361– 394, 2015

    work page 2015

  32. [32]

    Estimation of small failure probabilities in high dimensions by subset simulation,

    S.-K. Au and J. Beck, “Estimation of small failure probabilities in high dimensions by subset simulation,” Probabilistic Eng. Mech., vol. 16, no. 4, pp. 263–277, 2001

    work page 2001

  33. [33]

    Attention is all you need,

    A. Vaswani et al., “Attention is all you need,” in NeurIPS, vol. 30, Curran Associates, Inc., 2017

    work page 2017

  34. [34]

    Variational mixture-of-experts autoencoders for multi-modal deep generative models,

    Y. Shi et al., “Variational mixture-of-experts autoencoders for multi-modal deep generative models,” in NeurIPS, vol. 32, Curran Associates, Inc., 2019

    work page 2019

  35. [35]

    An interface setup optimization method using a throughput estimation model for concurrently communicat- ing access points in a wireless local area network,

    F. Akhter et al., “An interface setup optimization method using a throughput estimation model for concurrently communicat- ing access points in a wireless local area network,” Sensors, vol. 23, no. 14, p. 6367, 2023

    work page 2023

  36. [36]

    Denoising Diffusion Implicit Models

    J. Song et al., “Denoising diffusion implicit models,” arXiv preprint arXiv:2010.02502, 2020

    work page internal anchor Pith review arXiv 2010