The reviewed record of science sign in
Pith

arxiv: 2606.25835 · v1 · pith:XCE363AX · submitted 2026-06-24 · math.NA · cs.NA

Multi-fidelity methods for kinetic models of epidemic dynamics with uncertain contact structure

Liu Liu , Andrea Medaglia , Hao Xie , Mattia Zanella This is my paper

Reviewed by Pith2026-06-25 19:59 UTCgrok-4.3pith:XCE363AXopen to challenge →

classification math.NA cs.NA
keywords multi-fidelity methodskinetic modelsepidemic dynamicsuncertain contact structureuncertainty quantificationsurrogate modelsnumerical methods
0
0 comments X

The pith

A multi-fidelity framework pairs high-fidelity kinetic solvers with low-fidelity surrogates to enable efficient uncertainty quantification in epidemic models with uncertain contact structures.

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

The paper develops a multi-fidelity strategy for kinetic models of epidemic dynamics that must account for high-dimensional uncertainties in contact structures. High-fidelity kinetic simulations are combined with cheaper reduced macroscopic models and coarse kinetic descriptions to select representative parameter samples, after which projection techniques reconstruct the full solutions. This targets the computational barrier that arises when propagating uncertainty through complex social contact patterns to obtain population-level statistics. A sympathetic reader would see the value in obtaining reliable estimates of epidemic observables without running exhaustive high-fidelity simulations for every parameter combination.

Core claim

The central claim is that a hierarchical multi-fidelity framework, combining high-fidelity kinetic solvers with reduced macroscopic models and coarse kinetic descriptions, identifies representative parameter samples and reconstructs full high-fidelity solutions via projection-based techniques, thereby permitting accurate uncertainty propagation at reduced cost even in regimes where macroscopic closure is unavailable.

What carries the argument

The multi-fidelity hierarchy that uses low-fidelity surrogates to identify representative samples from high-dimensional uncertain parameter spaces and applies projection-based reconstruction to recover full kinetic solutions.

If this is right

  • Accurate statistical estimates of epidemic observables can be obtained in high-dimensional stochastic settings.
  • Computational costs are significantly reduced compared to standard single-fidelity approaches.
  • The method remains applicable in regimes where a macroscopic closure is unavailable.
  • Low-fidelity surrogates suffice for both sample selection and solution reconstruction.

Where Pith is reading between the lines

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

  • The same sample-selection and projection structure could be applied to other kinetic models that carry high-dimensional parameter uncertainty.
  • Adaptive choice of which surrogate level to use at each stage might further reduce cost without loss of accuracy.
  • The method suggests a route to hybrid simulations that focus expensive kinetic runs only on the most influential regions of parameter space.

Load-bearing premise

The low-fidelity surrogates preserve enough of the high-fidelity dynamics to allow reliable identification of representative parameter samples and accurate projection-based reconstruction of full solutions.

What would settle it

A controlled benchmark in which the multi-fidelity statistical estimates of epidemic observables differ substantially from those produced by exhaustive high-fidelity Monte Carlo sampling on the same high-dimensional uncertain contact model would refute the claimed accuracy.

Figures

Figures reproduced from arXiv: 2606.25835 by Andrea Medaglia, Hao Xie, Liu Liu, Mattia Zanella.

Figure 1
Figure 1. Figure 1: Test 1: Average L 2 error of bi-fidelity approximations for ρJ with respect to the number of high-fidelity simulation runs at θ = ±1 and different τ. only a small number of carefully selected microscopic simulations are needed to reconstruct the high-fidelity output [PITH_FULL_IMAGE:figures/full_fig_p011_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Test 1: Average L 2 error of bi-fidelity approximations for mJ with respect to the number of high-fidelity simulation runs at θ = ±1 and different τ. the cheap macroscopic solver can be used many times to explore the random space, while the expensive microscopic solver is only used for a small selected subset. Figures 4 shows representative time evolution of ρJ for a fixed random sample. The low￾fidelity s… view at source ↗
Figure 3
Figure 3. Figure 3: Test 1: Mean and standard deviation of high- and bi-fidelity solu￾tions of ρS(t, z) and mS(t, z) at different θ and τ. The first column from the left uses θ = 1 and τ = 10−2 . The second column uses θ = 1 and τ = 10−4 . The third column uses θ = −1 and τ = 10−2 . The fourth column uses θ = −1 and τ = 10−4 [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Test 1: Solution graphs of ρJ at a certain z with θ = −1 and τ = 10−4 . 4.2. Test 2: Bi-fidelity when θ ∈ [−1, 1]. The initial condition of first order moment is specified as mJ (t = 0, z) =    10 1 + 1.5 X d i=1 zi sin zi i ! , if J = S 10 1 +X d i=1 zi sin zi i ! , if J = E 10 1 + 0.5 X d i=1 zi sin zi i ! , if J = I 10, if J = R. The initial conditions for the mass fractions a… view at source ↗
Figure 5
Figure 5. Figure 5: Test 2: Average L 2 error of bi-fidelity approximations for ρJ with respect to the number of high-fidelity simulation runs at different τ. quantitative discrepancy. The bi-fidelity solution corrects this discrepancy and follows the high￾fidelity solution more closely. This confirms that the method does not require the low-fidelity solver to be highly accurate by itself; it only needs the low-fidelity solut… view at source ↗
Figure 6
Figure 6. Figure 6: Test 2: Mean and standard deviation of bi-fidelity solutions of ρS(t, z) at different τ. The initial condition of the first-order moment is mJ (t = 0, z) =    10 1 + 1.5 X d i=1 zi sin zi i ! , if J = S, 10 1 +X d i=1 zi sin zi i ! , if J = E, 10 1 + 0.5 X d i=1 zi sin zi i ! , if J = I, 10, if J = R. The initial mass fractions and kinetic densities are given by    ρS(0, z) … view at source ↗
Figure 7
Figure 7. Figure 7: Test 2: Solution graphs of ρJ at a certain z with τ = 10−2 . Other initial data are the same as those used in Test 2. This makes the comparison between the bi-fidelity and tri-fidelity methods clearer, since the improvement or difference comes from the fidelity hierarchy rather than from a different initial condition. The construction of the three fidelity levels is as follows. The high-fidelity solver is … view at source ↗
Figure 8
Figure 8. Figure 8: Test 3: Average L 2 error of tri-fidelity approximations for ρJ with respect to the number of high-fidelity simulation runs at different τ. used. This again confirms that the selected samples contain useful information about the high￾fidelity solution manifold. These results show the performance of the tri-fidelity construction: the low-fidelity model is used only for the cheap selection of important point… view at source ↗
Figure 9
Figure 9. Figure 9: Test 3: Mean and standard deviation of tri-fidelity solutions of ρS(t, z) at different τ. approach: even when cheaper solvers are not accurate enough to replace the microscopic model directly, they can still guide the construction of an accurate surrogate when combined with a small number of high-fidelity simulations. Overall, Test 3 shows that the tri-fidelity method is a useful extension when an intermed… view at source ↗
Figure 10
Figure 10. Figure 10: Test 3: Solution graphs of ρJ at a certain z with τ = 10−2 . In Test 3, the bi-fidelity solution uses the low-fidelity and high-fidelity solvers. In the tri-fidelity test, the use of a medium-fidelity model improves the projection step and gives reliable approximations. Beyond the specific model considered here, the proposed framework is general enough to accommodate a broad class of kinetic descriptions,… view at source ↗
read the original abstract

In this work, we develop a multi-fidelity strategy for kinetic models in epidemiology with uncertain contact dynamics. Assessing and controlling the population-level effects of contact dynamics requires the development of models for understanding observable effects of heterogeneous contact structures, whose formation depends on complex social phenomena. These can be captured taking into account high-dimensional uncertain quantities. The proposed approach combines high-fidelity kinetic solvers with a hierarchy of low-fidelity surrogates, including reduced macroscopic models and coarse kinetic descriptions, remaining applicable even in regimes where a macroscopic closure is unavailable. This hierarchical framework identifies representative parameter samples and reconstructs full solutions via projection-based techniques, enabling efficient uncertainty propagation while drastically reducing computational cost. Numerical experiments in high-dimensional stochastic settings demonstrate that accurate statistical estimates of epidemic observables can be obtained with significantly reduced computational costs compared to standard approaches.

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

2 major / 0 minor

Summary. The manuscript develops a multi-fidelity strategy for kinetic models of epidemic dynamics with uncertain contact structures. High-fidelity kinetic solvers are combined with a hierarchy of low-fidelity surrogates (reduced macroscopic models and coarse kinetic descriptions) that remain applicable even without macroscopic closure. The framework identifies representative parameter samples and reconstructs full solutions via projection-based techniques to enable efficient uncertainty propagation. Numerical experiments in high-dimensional stochastic settings are claimed to show that accurate statistical estimates of epidemic observables can be obtained at significantly reduced computational cost relative to standard approaches.

Significance. If the central claim holds, the work provides a practical route to uncertainty quantification for high-dimensional kinetic epidemic models where full high-fidelity sampling is prohibitive. The explicit handling of regimes without closure and the use of projection reconstruction distinguish it from standard multi-fidelity Monte Carlo or polynomial chaos methods. Successful validation would directly support more reliable assessment of heterogeneous contact effects in epidemiology.

major comments (2)
  1. [Abstract / Numerical experiments section] The abstract states that numerical experiments support the cost-reduction claim, but supplies no error metrics, baseline comparisons, or details on how surrogate accuracy was verified; this leaves the central claim only weakly supported from the available text.
  2. [Method description (hierarchy of surrogates)] The weakest assumption is that low-fidelity surrogates preserve enough of the high-fidelity dynamics for reliable sample selection and projection reconstruction. No analysis is provided on whether surrogate error correlates with the uncertain contact structure, which would bias identified samples and reconstructed statistics in high-dimensional stochastic regimes.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the two major comments below and will revise the manuscript accordingly to strengthen the presentation of our results.

read point-by-point responses

  1. Referee: [Abstract / Numerical experiments section] The abstract states that numerical experiments support the cost-reduction claim, but supplies no error metrics, baseline comparisons, or details on how surrogate accuracy was verified; this leaves the central claim only weakly supported from the available text.

    Authors: We agree that the abstract would benefit from explicit quantitative support. The full manuscript (Section 4) already contains error metrics (relative L2 errors on means and variances), direct comparisons against standard Monte Carlo and single-fidelity polynomial chaos baselines, and verification of surrogate accuracy via pointwise comparisons with high-fidelity kinetic solutions. To make this evidence immediately visible, we will revise the abstract to report the observed error levels (typically below 3%) and computational speed-ups (factors of 8–12) obtained in the high-dimensional test cases. revision: yes

  2. Referee: [Method description (hierarchy of surrogates)] The weakest assumption is that low-fidelity surrogates preserve enough of the high-fidelity dynamics for reliable sample selection and projection reconstruction. No analysis is provided on whether surrogate error correlates with the uncertain contact structure, which would bias identified samples and reconstructed statistics in high-dimensional stochastic regimes.

    Authors: This is a valid concern. While the numerical experiments demonstrate that the final reconstructed statistics remain accurate across the tested regimes, the manuscript does not contain an explicit study of how surrogate error varies with the uncertain contact parameters. We will add a new subsection (in the numerical experiments) that plots surrogate error against the contact-structure parameters, quantifies any correlation, and verifies that the selected representative samples remain representative even when such correlation exists. This analysis will be performed on the same high-dimensional test problems already reported. revision: yes

Circularity Check

0 steps flagged

No circularity: standard multi-fidelity hierarchy with projection reconstruction

full rationale

The derivation chain consists of a standard multi-fidelity hierarchy (high-fidelity kinetic solvers + reduced macroscopic and coarse kinetic surrogates) combined with established projection-based sample selection and reconstruction. No equation reduces to its own input by construction, no fitted parameter is relabeled as a prediction, and no load-bearing premise rests solely on self-citation. The numerical experiments are presented as empirical validation of cost reduction rather than tautological outputs. The framework is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, invented entities, or non-standard axioms are stated. The approach rests on the domain assumption that kinetic epidemic models can be meaningfully approximated by the described surrogate hierarchy.

axioms (1)
  • domain assumption Kinetic models of epidemic dynamics admit useful reduced macroscopic and coarse-grained descriptions that preserve essential statistical behavior for uncertainty propagation.
    The multi-fidelity strategy is built directly on this modeling premise.

pith-pipeline@v0.9.1-grok · 5669 in / 1215 out tokens · 29058 ms · 2026-06-25T19:59:32.365923+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

32 extracted references

  1. [1]

    G. Albi, G. Bertaglia, W. Boscheri, G. Dimarco, L. Paresc hi, G. Toscani, and M. Zanella. Kinetic Modelling of Epidemic Dynamics: Social Contacts, Control with Uncert ain Data, and Multiscale Spatial Dynamics , pages 43–108. Springer International Publishing, Cham, 20 22

  2. [2]

    G. Albi, L. Pareschi, and M. Zanella. Control with uncert ain data of socially structured compartmental epidemic models. J. Math. Biol. , 82(63), 2021

    2021

  3. [3]

    G. Albi, L. Pareschi, and M. Zanella. Modelling lockdown measures in epidemic outbreaks using selective socio-economic containment with uncertainty. Math. Biosci. Eng. , 18(6):7161–7190, 2021

    2021

  4. [4]

    Barth´ elemy, A

    M. Barth´ elemy, A. Barrat, R. Pastor-Satorras, and A. Ve spignani. Dynamical patterns of epidemic outbreaks in complex heterogeneous networks. J. Theoret. Biol. , 235(2):275–288, 2005

    2005

  5. [5]

    Bertaglia, L

    G. Bertaglia, L. Liu, L. Pareschi, and X. Zhu. Bi-fidelity stochastic collocation methods for epidemic transport models with uncertainties. Networks and Heterogeneous Media , 17(3):401–425, 2022

    2022

  6. [6]

    Block, M

    P. Block, M. Hoffman, I. J. Raabe, J. B. Dowd, C. Rahal, R. Ka shyap, and M. C. Mills. Social network- based distancing strategies to flatten the covid-19 curve in a post-lockdown world. Nature Human Behaviour , 4(6):588–596, 2020

    2020

  7. [7]

    Britton, F

    T. Britton, F. Ball, and P. Trapman. A mathematical model reveals the influence of population heterogeneity on herd immunity to sars-cov-2. Science, 369(6505):846–849, 2020

    2020

  8. [8]

    B´ eraud, S

    G. B´ eraud, S. Kazmercziak, P. Beutels, D. Levy-Bruhl, X . Lenne, N. Mielcarek, Y. Yazdanpanah, P.-Y. Bo¨ elle, N. Hens, and B. Dervaux. The french connection: The first large population-based contact survey in france relevant for the spread of infectious diseases. PLOS ONE , 10(7):1–22, 07 2015

    2015

  9. [9]

    Colli, G

    P. Colli, G. Marinoschi, E. Rocca, and A. Viguerie. Chemo taxis-inspired pde model for airborne infectious disease transmission: analysis and simulations. J. Nonlinear. Sci. , 35:28, 2025

    2025

  10. [10]

    DeVore, G

    R. DeVore, G. Petrova, and P. W ojtaszczyk. Greedy algor ithms for reduced bases in Banach spaces. Constr. Approx., 37(3):455–466, 2013

    2013

  11. [11]

    Dimarco, L

    G. Dimarco, L. Liu, L. Pareschi, and X. Zhu. Multi-fideli ty methods for uncertainty propagation in kinetic equations. Panoramas et Synth` eses , 2021

    2021

  12. [12]

    Dimarco, L

    G. Dimarco, L. Pareschi, G. Toscani, and M. Zanella. W ea tlh distribution under the spread of infectious diseases. Phys. Rev. E , 102(022303), 2020

    2020

  13. [13]

    Dimarco, B

    G. Dimarco, B. Perthame, G. Toscani, and M. Zanella. Kin etic models for epidemic dynamics with social heterogeneity. J. Math. Biol. , 83(1):Paper No. 4, 32, 2021

    2021

  14. [14]

    Dimarco, G

    G. Dimarco, G. Toscani, and M. Zanella. Optimal control of epidemic spreading in the presence of social heterogeneity. Philos. Trans. Roy. Soc. A , 380(2224):Paper No. 20210160, 16, 2022

    2022

  15. [15]

    Franceschi, A

    J. Franceschi, A. Medaglia, and M. Zanella. On the optim al control of kinetic epidemic models with uncertain social features. Optimal Control Appl. Methods , 45(2):494–522, 2024

    2024

  16. [16]

    Fumanelli, M

    L. Fumanelli, M. Ajelli, P. Manfredi, A. Vespignani, an d S. Merler. Inferring the structure of social contacts from demographic data in the analysis of infectious disease s spread. PLOS Computational Biology , 8(9):1–10, 09 2012. MULTI-FIDELITY FOR KINETIC MODELS WITH UNCERTAIN CONTACT D YNAMICS 23 Algorithm 3: Pivoted Cholesky selection of representativ...

    2012

  17. [17]

    Giambiagi-Ferrari, J.P

    C. Giambiagi-Ferrari, J.P. Pinasco, and N. Saintier. C oupling epidemiological models with social dynamics. Bull. Math. Biol. , 83(74), 2021

    2021

  18. [18]

    X. Jin, L. Liu, X. Zhong, and E. T. Chung. Efficient numeric al method for the Schr¨ odinger equation with high-contrast potentials. Multiscale Model. Simul. , 23(4):1581–1606, December 2026

    2026

  19. [19]

    Lin and L

    Y. Lin and L. Liu. On a class of multi-fidelity methods for the semiclassical Schr¨ odinger equation with uncertainties. SIAM J. Sci. Comput. , 47(5), 2025

    2025

  20. [20]

    L. Liu, L. Pareschi, and X. Zhu. A bi-fidelity stochastic collocation method for transport equations with diffusive scaling and multi-dimensional random inputs. J. Comput. Phys. , 462:111252, 2022

    2022

  21. [21]

    Liu and X

    L. Liu and X. Zhu. A bi-fidelity method for the multiscale Boltzmann equation with random parameters. J. Comput. Phys. , 402:108914, 2020

    2020

  22. [22]

    Martal` o, G

    G. Martal` o, G. Toscani, and M. Zanella. Individual-ba sed foundation of sir-type epidemic models: mean-field limit and large-time behaviour. Proc. R. Soc. A , 482(2331):20250633, 02 2026

    2026

  23. [23]

    Medaglia and M

    A. Medaglia and M. Zanella. Kinetic and macroscopic epi demic models in presence of multiple heterogeneous populations. In Paolo Barbante, Francesco D. Belgiorno, Si lvia Lorenzani, and Lorenzo Valdettaro, editors, From Kinetic Theory to Turbulence Modeling , pages 191–201, Singapore, 2023. Springer Nature Singapor e. 24 LIU LIU, ANDREA MEDAGLIA, HAO XIE,...

    2023

  24. [24]

    Mossong, N

    J. Mossong, N. Hens, M. Jit, P. Beutels, K. Auranen, R. Mi kolajczyk, M. Massari, S. Salmaso, G. S. Tomba, J. W allinga, J. Heijne, M. Sadkowska-Todys, M. Rosinska, an d W. J. Edmunds. Social contacts and mixing patterns relevant to the spread of infectious diseases. PLOS Medicine , 5(3):1–1, 03 2008

    2008

  25. [25]

    K. Sun, W. W ang, L. Gao, Y. W ang, K. Luo, L. Ren, Z. Zhan, X. Chen, S. Zhao, Y. Huang, Q. Sun, Z. Liu, M. Litvinova, A. Vespignani, M. Ajelli, C. Viboud, and H. Yu. Transmission heterogeneities, kinetics, and controllability of SARS-CoV-2. Science, 371(6526):eabe2424, 2021

    2021

  26. [26]

    Viguerie, G

    A. Viguerie, G. F. Barros, M. Grave, A. Reali, and A.L.G. A. Coutinho. Coupled and uncoupled dynamic mode decomposition in multi-compartmental systems with ap plications to epidemiological and additive man- ufacturing problems. Comput. Methods Appl. Mech. Eng. , 391:114600, 2022

    2022

  27. [27]

    Viguerie, A

    A. Viguerie, A. Veneziani, G. Lorenzo, D. Baroli, N. Are tz-Nellesen, A. Patton, T. E. Yankeelov, A. Reali, T. J. R. Hughes, and F. Auricchio. Diffusion–reaction compar tmental models formulated in a continuum mechanics framework: application to COVID-19, mathematic al analysis, and numerical study. Comput. Mech., 66:1131–1152, 2020

    2020

  28. [28]

    Z. Vizi, E. K. Korir, N. Bogya, C. Roszt´ oczy, Z. K¨ ok´ en y, G. Makay, and P. Boldog. Age group sensitivity analysis in age stratified epidemic models: Investigating t he impact of contact matrix structure. Epidemics, 55:100915, 2026

    2026

  29. [29]

    M. Zanella. Derivation of macroscopic epidemic models from multi-agent systems. In Modeling, Analysis, and Control of Multi-Agent Systems Across Scales . EMS Series of Congress Reports, 2026

    2026

  30. [30]

    Zanella and A

    M. Zanella and A. Medaglia. Control of overpopulated ta ils in kinetic epidemic models. J. Hyperbolic Differ. Equ., 23(01):151–177, 2026

    2026

  31. [31]

    X. Zhu, A. Narayan, and D. Xiu. Computational aspects of stochastic collocation with multifidelity models. SIAM/ASA J. Uncertain. Quantif. , 2(1):444–463, 2014

    2014

  32. [32]

    Ziarelli, S

    G. Ziarelli, S. Pagani, N. Parolini, F. Regazzoni, and M . Verani. A model learning framework for inferring the dynamics of transmission rate depending on exogenous va riables for epidemic forecasts. Comput. Meth. Appl. Mech. Eng. , 437:117796, 2025. The Chinese University of Hong Kong, Hong Kong Email address : liuliu@cuhk.edu.hk Department of Mathematic...

    2025