pith. sign in

arxiv: 2606.27286 · v1 · pith:MAHFXDKQnew · submitted 2026-06-25 · 💻 cs.AI

Simulation-based inference for rapid Bayesian parameter estimation in epidemiological models: a comparison with MCMC

Alina Bazarova , Johann Fredrik Jadebeck , Henrik Zunker , Carolina J. Klett-Tammen , Torben Heinsohn , Wolfgang Wiechert , Katharina Noeh , Stefan Kesselheim This is my paper

Pith reviewed 2026-06-26 04:24 UTC · model grok-4.3

classification 💻 cs.AI
keywords simulation-based inferenceneural posterior estimationMCMCSECIR modelBayesian calibrationepidemiological modelsCOVID-19ICU data
0
0 comments X

The pith

Simulation-based inference recovers MCMC posteriors for an SECIR epidemiological model while requiring far less computation.

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

This paper tests whether simulation-based inference with neural posterior estimation can replace Markov chain Monte Carlo for Bayesian calibration of a mechanistic SECIR model to German COVID-19 ICU occupancy data. It demonstrates that the two methods yield closely matching posterior distributions and accurate trajectory predictions across short and long time windows. The central benefit is speed, as SBI finishes in tens to low hundreds of seconds while MCMC takes thousands to tens of thousands of seconds. This difference matters for applications that require frequent recalibration during an evolving outbreak. The comparison covers both 31-day inference windows and a 201-day reconstruction with multiple transmission change points.

Core claim

Neural posterior estimation recovers posterior distributions for SECIR model parameters that show strong quantitative agreement with MCMC results, measured by Wasserstein distances and Kullback-Leibler divergences, and that produce posterior predictive checks matching observed ICU trajectories; the agreement holds for both 31-day and 201-day periods, with SBI preserving the main posterior features even under higher uncertainty in the longer window.

What carries the argument

Neural posterior estimation, which trains a neural network on simulated data from the SECIR model to approximate the posterior over parameters given ICU observations.

If this is right

  • SBI supports repeated near-real-time Bayesian analyses for infectious disease forecasting.
  • Computational runtime drops by factors of 15 to over 100 compared to MCMC on the tested problems.
  • The method maintains performance when extending to longer time series with multiple change points.
  • Posterior predictive checks confirm that inferred parameters reproduce the observed ICU occupancy data.

Where Pith is reading between the lines

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

  • The speed advantage could allow ensemble forecasting or integration with additional data streams in public health applications.
  • Similar gains may appear when applying the approach to other mechanistic models or different observation types.
  • Further tests on data with known ground-truth parameters would quantify any residual approximation error in the neural network.

Load-bearing premise

The trained neural network provides an accurate enough approximation to the true posterior for the observed ICU data in the tested time windows.

What would settle it

Finding a dataset or time window where SBI posteriors show large discrepancies from MCMC posteriors in Wasserstein distance or where the predicted trajectories deviate substantially from observed ICU occupancy.

Figures

Figures reproduced from arXiv: 2606.27286 by Alina Bazarova, Carolina J. Klett-Tammen, Henrik Zunker, Johann Fredrik Jadebeck, Katharina Noeh, Stefan Kesselheim, Torben Heinsohn, Wolfgang Wiechert.

Figure 1
Figure 1. Figure 1: Overview of the SECIR model compartment and transition structure. [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Workflow of amortized neural posterior estimation (NPE). [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Posterior predictive checks for the inferred SBI posterior distributions [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of marginal posterior distributions inferred using [PITH_FULL_IMAGE:figures/full_fig_p016_4.png] view at source ↗
read the original abstract

Mechanistic epidemiological models are widely used to support infectious disease forecasting and public-health decision making. Bayesian calibration of such models is commonly performed using Markov chain Monte Carlo (MCMC), which can become computationally expensive for high-dimensional nonlinear systems and repeated near-real-time analyses. Here, we investigate simulation-based inference (SBI) using neural posterior estimation as a scalable alternative for Bayesian calibration of a mechanistic SECIR epidemiological model using COVID-19 intensive care unit (ICU) occupancy data from Germany during 2020. We compared SBI and MCMC across multiple epidemic phases using both 31-day inference windows and a substantially more challenging 201-day reconstruction problem involving multiple transmission change points. Posterior agreement was evaluated quantitatively using Wasserstein distances and Kullback-Leibler divergences together with posterior predictive checks. Across the 31-day windows, SBI recovered posterior distributions in strong agreement with MCMC while accurately reproducing observed ICU trajectories. In the 201-day setting, SBI preserved the dominant posterior structure despite increased uncertainty. SBI, by combining CPU and GPU resources, substantially reduced computational runtime compared with MCMC, which was restricted to running on CPUs. Whereas MCMC required approximately 1000 seconds for the 31-day inference problems, SBI achieved comparable posterior and predictive performance in approximately 60-70 seconds on a single GPU. For the 201-day inference problem, SBI required an average of 157 seconds, while the MCMC runs took over 19,000 seconds. Our results demonstrate that SBI provides a rapid and computationally efficient framework for Bayesian calibration of mechanistic epidemiological models, supporting repeated near-real-time inference and rapid outbreak analysis.

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 / 2 minor

Summary. The manuscript compares simulation-based inference (SBI) via neural posterior estimation against MCMC for Bayesian calibration of a SECIR epidemiological model to German COVID-19 ICU occupancy data. It reports quantitative agreement between SBI and MCMC posteriors (via Wasserstein distances, KL divergences, and posterior predictive checks) across 31-day windows and a 201-day multi-change-point reconstruction, while claiming SBI achieves comparable results in far less time (60-70 s vs ~1000 s for 31-day; 157 s vs >19000 s for 201-day).

Significance. If the reported posterior agreement holds without substantial approximation bias, the work supplies a concrete, scalable alternative to MCMC for repeated Bayesian calibration of nonlinear mechanistic models. The runtime numbers and agreement metrics on real ICU data would be directly useful for near-real-time epidemiological applications.

major comments (2)
  1. [abstract and §5] The central claim that SBI recovers posteriors in strong agreement with MCMC for the 201-day multi-change-point case (abstract and §5) rests on the unverified assumption that the 10^5–10^6 training simulations densely cover the relevant region of parameter space. No prior-predictive coverage diagnostics, amortized posterior calibration on held-out simulations, or effective-sample-size comparisons are reported that would confirm this coverage for the transmission change-point parameters.
  2. [§4.2] §4.2 and the methods description of the NPE network provide no quantitative assessment of approximation bias (e.g., via simulation-based calibration or posterior coverage probabilities) that would be required to interpret the reported Wasserstein and KL distances as evidence of faithful posterior recovery rather than under-sampling artifacts.
minor comments (2)
  1. [abstract and §5] The hardware specifications and parallelization details for the MCMC runs (beyond “restricted to CPUs”) are not stated, making the runtime comparison in the abstract and §5 difficult to interpret.
  2. [§2] Notation for the SECIR parameters and change-point priors is introduced without a consolidated table; a single reference table would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which correctly identify opportunities to strengthen the validation of the SBI approximation. We address each major comment below.

read point-by-point responses

  1. Referee: [abstract and §5] The central claim that SBI recovers posteriors in strong agreement with MCMC for the 201-day multi-change-point case (abstract and §5) rests on the unverified assumption that the 10^5–10^6 training simulations densely cover the relevant region of parameter space. No prior-predictive coverage diagnostics, amortized posterior calibration on held-out simulations, or effective-sample-size comparisons are reported that would confirm this coverage for the transmission change-point parameters.

    Authors: We agree that the manuscript would benefit from explicit coverage diagnostics for the 201-day case. The reported agreement with MCMC (Wasserstein distances, KL divergences, and posterior predictive checks) provides supporting evidence that the training simulations were sufficient, but this is indirect. In the revised version we will add prior-predictive coverage diagnostics and simulation-based calibration on held-out simulations, with particular attention to the change-point parameters. revision: yes

  2. Referee: [§4.2] §4.2 and the methods description of the NPE network provide no quantitative assessment of approximation bias (e.g., via simulation-based calibration or posterior coverage probabilities) that would be required to interpret the reported Wasserstein and KL distances as evidence of faithful posterior recovery rather than under-sampling artifacts.

    Authors: We acknowledge the absence of direct approximation-bias diagnostics in the current text. The manuscript currently uses agreement with MCMC as the primary validation. We will incorporate quantitative assessments of approximation bias, including simulation-based calibration and posterior coverage probabilities, into §4.2 and the methods description in the revision. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical benchmark comparison of SBI vs MCMC

full rationale

The paper reports runtime and posterior-agreement metrics (Wasserstein, KL, predictive checks) between two independent inference methods applied to the same SECIR model and ICU data. MCMC is treated as an external reference, not derived from SBI outputs or vice versa. No equations reduce a claimed result to a fitted parameter by construction, no self-citation chain supports a uniqueness claim, and no ansatz or renaming is presented as a derivation. The work is a self-contained empirical study whose central claims rest on observable performance differences rather than internal re-derivation.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Abstract-only review limits visibility into exact assumptions; central claim rests on standard Bayesian simulation assumptions plus the adequacy of the SECIR model for the 2020 German COVID data.

axioms (1)
  • domain assumption The SECIR mechanistic model is an adequate representation of COVID-19 transmission dynamics in Germany during 2020
    Used as the forward simulator for both SBI and MCMC calibration against ICU occupancy.

pith-pipeline@v0.9.1-grok · 5850 in / 1286 out tokens · 65119 ms · 2026-06-26T04:24:55.091235+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

50 extracted references · 23 canonical work pages

  1. [1]

    Compart- mental structures used in modeling covid-19: a scoping review.Infectious Diseases of Poverty, 11(1):72, Jun 2022

    Lingcai Kong, Mengwei Duan, Jin Shi, Jie Hong, Zhaorui Chang, and Zhijie Zhang. Compart- mental structures used in modeling covid-19: a scoping review.Infectious Diseases of Poverty, 11(1):72, Jun 2022. ISSN 2049-9957. doi: 10.1186/s40249-022-01001-y

    work page doi:10.1186/s40249-022-01001-y 2022

  2. [2]

    Eduard Campillo-Funollet, James Van Yperen, Phil Allman, Michael Bell, Warren Beresford, Jacqueline Clay, Matthew Dorey, Graham Evans, Kate Gilchrist, Anjum Memon, Gurprit Pannu, Ryan Walkley, Mark Watson, and Anotida Madzvamuse. Predicting and forecasting the impact of local outbreaks of covid-19: use of seir-d quantitative epidemiological modelling for ...

    work page doi:10.1093/ije/dyab106 2021

  3. [3]

    Differential effects of intervention timing on covid-19 spread in the united states.Science Advances, 6(49):eabd6370, 2020

    Sen Pei, Sasikiran Kandula, and Jeffrey Shaman. Differential effects of intervention timing on covid-19 spread in the united states.Science Advances, 6(49):eabd6370, 2020. doi: 10.1126/ sciadv.abd6370

    2020

  4. [4]

    Jonas Gilg, Johann F. Jadebeck, Mariama Jaiteh, David Kerkmann, Niklas Medinger, Shabaz Memon, Anna Clara Wendler, Moritz Zeumer, Henrik Zunker, Maximilian Franz Betz, Ralf Hannemann-Tamas, Jonas Heinicke, Julian Litz, Achim Basermann, Cas Cremers, Manuel Dahmen, Andreas Gerndt, Jens Henrik G¨ obbert, Bj¨ orn Hagemeier, Carolina J. Klett-Tammen, Berit Lan...

    2026

  5. [5]

    Paul Spitzner, Michael Wibral, Joao Pinheiro Neto, Michael Wilczek, and Viola Priesemann

    Jonas Dehning, Johannes Zierenberg, F. Paul Spitzner, Michael Wibral, Joao Pinheiro Neto, Michael Wilczek, and Viola Priesemann. Inferring change points in the spread of covid-19 reveals the effectiveness of interventions.Science, 369(6500):eabb9789, 2020. doi: 10.1126/ science.abb9789

    2020

  6. [6]

    Juliette T

    Seth Flaxman, Swapnil Mishra, Axel Gandy, H. Juliette T. Unwin, Thomas A. Mellan, Helen Coupland, Charles Whittaker, Harrison Zhu, Tresnia Berah, Jeffrey W. Eaton, M´ elodie Monod, Pablo N. Perez-Guzman, Nora Schmit, Lucia Cilloni, Kylie E. C. Ainslie, Marc Baguelin, Adhiratha Boonyasiri, Olivia Boyd, Lorenzo Cattarino, Laura V. Cooper, Zulma Cucunub´ a, ...

    work page doi:10.1038/s41586-020-2405-7 2020

  7. [7]

    Kucharski, Rosalind M

    Sebastian Funk, Anton Camacho, Adam J. Kucharski, Rosalind M. Eggo, and W. John Edmunds. Real-time forecasting of infectious disease dynamics with a stochastic semi-mechanistic model. Epidemics, 22:56–61, 2018. ISSN 1755-4365. doi: 10.1016/j.epidem.2016.11.003. The RAPIDD Ebola Forecasting Challenge

    work page doi:10.1016/j.epidem.2016.11.003 2018

  8. [8]

    The rapidd ebola forecasting challenge: Synthesis and lessons learnt.Epidemics, 22:13–21, 2018

    C´ ecile Viboud, Kaiyuan Sun, Robert Gaffey, Marco Ajelli, Laura Fumanelli, Stefano Merler, Qian Zhang, Gerardo Chowell, Lone Simonsen, and Alessandro Vespignani. The rapidd ebola forecasting challenge: Synthesis and lessons learnt.Epidemics, 22:13–21, 2018. ISSN 1755-4365. doi: 10.1016/j.epidem.2017.08.002. The RAPIDD Ebola Forecasting Challenge

    work page doi:10.1016/j.epidem.2017.08.002 2018

  9. [9]

    A collaborative multiyear, multimodel assessment of seasonal influenza forecasting in the United States,

    Nicholas G. Reich, Logan C. Brooks, Spencer J. Fox, Sasikiran Kandula, Craig J. McGowan, Evan Moore, Dave Osthus, Evan L. Ray, Abhinav Tushar, Teresa K. Yamana, Matthew Bigger- staff, Michael A. Johansson, Roni Rosenfeld, and Jeffrey Shaman. A collaborative multiyear, multimodel assessment of seasonal influenza forecasting in the united states.Proceedings...

    work page doi:10.1073/pnas.1812594116 2019

  10. [10]

    Craiu and Jeffrey S

    Radu V. Craiu and Jeffrey S. Rosenthal. Bayesian computation via markov chain monte carlo.Annual Review of Statistics and Its Application, 1(Volume 1, 2014):179–201, 2014. ISSN 2326-831X. doi: 10.1146/annurev-statistics-022513-115540

    work page doi:10.1146/annurev-statistics-022513-115540 2014

  11. [11]

    Jewell, Theodoros Kypraios, Paul Neal, and Gareth O

    Nicholas P. Jewell, Theodoros Kypraios, Paul Neal, and Gareth O. Roberts. Bayesian analysis for emerging infectious diseases.Journal of the Royal Statistical Society: Series C (Applied Statistics), 58(2):317–336, 2009

    2009

  12. [12]

    Fitting mechanistic epidemic models to data: A comparison of simple markov chain monte carlo approaches.Stat Methods Med Res, 27(7):1956–1967, July 2018

    Michael Li, Jonathan Dushoff, and Benjamin M Bolker. Fitting mechanistic epidemic models to data: A comparison of simple markov chain monte carlo approaches.Stat Methods Med Res, 27(7):1956–1967, July 2018

    1956

  13. [13]

    Bayesian uncertainty quantification for transmissibility of influenza, norovirus and ebola using information geometry.J R Soc Interface, 13(121), August 2016

    Thomas House, Ashley Ford, Shiwei Lan, Samuel Bilson, Elizabeth Buckingham-Jeffery, and Mark Girolami. Bayesian uncertainty quantification for transmissibility of influenza, norovirus and ebola using information geometry.J R Soc Interface, 13(121), August 2016

    2016

  14. [14]

    Craiu and Jeffrey S

    Radu V. Craiu and Jeffrey S. Rosenthal. Bayesian computation via markov chain monte carlo.Annual Review of Statistics and Its Application, 1:179–201, 2014. doi: 10.1146/ annurev-statistics-022513-115540

    2014

  15. [15]

    Integrative modelling of reported case numbers and seroprevalence reveals time-dependent test efficiency and infectious contacts

    Lorenzo Contento, Noemi Castelletti, Elba Raim´ undez, Ronan Le Gleut, Yannik Sch¨ alte, Paul Stapor, Ludwig Christian Hinske, Michael Hoelscher, Andreas Wieser, Katja Radon, Christiane Fuchs, Jan Hasenauer, and the KoCo19 study group. Integrative modelling of reported case numbers and seroprevalence reveals time-dependent test efficiency and infectious c...

    work page doi:10.1016/j.epidem.2023.100681 2023

  16. [16]

    The frontier of simulation-based inference

    Kyle Cranmer, Johann Brehmer, and Gilles Louppe. The frontier of simulation-based inference. Proceedings of the National Academy of Sciences, 117(48):30055–30062, 2020. doi: 10.1073/ pnas.1912789117

    2020

  17. [17]

    Greenberg, Pedro J

    Jan-Matthis Lueckmann, Jan Boelts, David S. Greenberg, Pedro J. Gon¸ calves, and Jakob H. Macke. Benchmarking simulation-based inference, 2021. 22

    2021

  18. [18]

    Flexible bayesian inference on partially observed epidemics.J Complex Netw, 12(2):cnae017, March 2024

    Maxwell H Wang and Jukka-Pekka Onnela. Flexible bayesian inference on partially observed epidemics.J Complex Netw, 12(2):cnae017, March 2024

    2024

  19. [19]

    Simulation based- inference of epidemiological and phylodynamic models via neural posterior estimation.bioRxiv,

    Francesco Pinotti, Julien Th´ ez´ e, Xavier Bailly, and Guillaume Fourni´ e. Simulation based- inference of epidemiological and phylodynamic models via neural posterior estimation.bioRxiv,

  20. [20]

    doi: 10.1101/2025.11.25.690436

    work page doi:10.1101/2025.11.25.690436 2025

  21. [21]

    Assessment of simulation-based inference methods for stochastic compartmental models in epidemiological research, 2026

    Vincent Wieland, Nils Wassmuth, Lorenzo Contento, Martin K¨ uhn, and Jan Hasenauer. Assessment of simulation-based inference methods for stochastic compartmental models in epidemiological research, 2026

    2026

  22. [22]

    Manzano-Patr´ on, Michael Deistler, Cornelius Schr¨ oder, Theodore Kypraios, Pedro J

    J.P. Manzano-Patr´ on, Michael Deistler, Cornelius Schr¨ oder, Theodore Kypraios, Pedro J. Gon¸ calves, Jakob H. Macke, and Stamatios N. Sotiropoulos. Uncertainty mapping and probabilistic tractography using simulation-based inference in diffusion mri: A compari- son with classical bayes.Medical Image Analysis, 103:103580, 2025. ISSN 1361-8415. doi: 10.10...

    work page doi:10.1016/j.media.2025.103580 2025

  23. [23]

    Martin J. K¨ uhn, Daniel Abele, Tanmay Mitra, Wadim Koslow, Majid Abedi, Kathrin Rack, Martin Siggel, Sahamoddin Khailaie, Margrit Klitz, Sebastian Binder, Luca Spataro, Jonas Gilg, Jan Kleinert, Matthias H¨ aberle, Lena Pl¨ otzke, Christoph D. Spinner, Melanie Stecher, Xiao Xiang Zhu, Achim Basermann, and Michael Meyer-Hermann. Assessment of effective mi...

    work page doi:10.1016/j.mbs.2021.108648 2021

  24. [24]

    Julia Bicker, Carlotta Gerstein, David Kerkmann, Sascha Korf, Ren´ e Schmieding, Anna Wendler, Henrik Zunker, Daniel Abele, Maximilian Betz, Khoa Nguyen, Lena Pl¨ otzke, Kilian Volmer, Agatha Schmidt, Nils Waßmuth, Patrick Lenz, Daniel Richter, Hannah Tritzschak, Ralf Hannemann-Tamas, Julian Litz, Paul Johannssen, Marielena Borges, Annika Jungklaus, Manue...

    2026

  25. [25]

    Intensivkapazit¨ aten und covid-19-intensivbettenbelegung in deutschland, November 2025

    Robert Koch-Institut. Intensivkapazit¨ aten und covid-19-intensivbettenbelegung in deutschland, November 2025

    2025

  26. [26]

    Hendrik Streeck, Bianca Schulte, Beate M. K¨ ummerer, Enrico Richter, Tobias H¨ oller, Christine Fuhrmann, Eva Bartok, Ramona Dolscheid-Pommerich, Moritz Berger, Lukas Wessendorf, Monika Eschbach-Bludau, Angelika Kellings, Astrid Schwaiger, Martin Coenen, Per Hoffmann, Birgit Stoffel-Wagner, Markus M. N¨ othen, Anna M. Eis-H¨ ubinger, Martin Exner, Ricard...

    work page doi:10.1038/s41467-020-19509-y 2020

  27. [27]

    Sars-cov-2 seroprevalence in germany.Deutsches ¨Arzteblatt international, December 2021

    Daniela Gornyk, Manuela Harries, Stephan Gl¨ ockner, Monika Strengert, Tobias Kerrinnes, Jana- Kristin Heise, Henrike Maaß, Julia Ortmann, Barbora Kessel, Yvonne Kemmling, Berit Lange, and G´ erard Krause. Sars-cov-2 seroprevalence in germany.Deutsches ¨Arzteblatt international, December 2021. ISSN 1866-0452. doi: 10.3238/arztebl.m2021.0364. 23

    work page doi:10.3238/arztebl.m2021.0364 2021

  28. [28]

    Henrik Zunker, Ren´ e Schmieding, David Kerkmann, Alain Schengen, Sophie Diexer, Rafael Mikolajczyk, Michael Meyer-Hermann, and Martin J. K¨ uhn. Novel travel time aware metapop- ulation models and multi-layer waning immunity for late-phase epidemic and endemic scenarios. PLOS Computational Biology, 20(12), Dez 2024. doi: 10.1371/journal.pcbi.1012630

    work page doi:10.1371/journal.pcbi.1012630 2024

  29. [29]

    Sars-cov-2 infektionen in deutschland, 2025

    Robert Koch-Institut. Sars-cov-2 infektionen in deutschland, 2025

    2025

  30. [30]

    Fortschreibung des Bev¨ olkerungsstandes: 12411-02-03-4 Bev¨ olkerung nach Geschlecht und Altersgruppen (17) - Stichtag 31.12

    Regionaldatenbank Deutschland. Fortschreibung des Bev¨ olkerungsstandes: 12411-02-03-4 Bev¨ olkerung nach Geschlecht und Altersgruppen (17) - Stichtag 31.12. - regionale Tiefe: Kreise und krfr. St¨ adte, 2022. URLhttps://www.regionalstatistik.de/genesis/online? operation=statistic&levelindex=0&levelid=1646144362683&code=12411#abreadcrumb

    2022

  31. [31]

    Krankheitsschwere der ersten COVID-19-welle in deutschland basierend auf den meldungen gem¨ aß infektionsschutzgesetz

    Julia Schilling, Ann-Sophie Lehfeld, Dirk Schumacher, Alexander Ullrich, Michaela Diercke, Silke Buda, Walter Haas, and RKI COVID-19 Study Group. Krankheitsschwere der ersten COVID-19-welle in deutschland basierend auf den meldungen gem¨ aß infektionsschutzgesetz. Journal of Health Monitoring, 5(S11):2–20, 2020. doi: 10.25646/7169

    work page doi:10.25646/7169 2020

  32. [32]

    Oyungerel Byambasuren, Magnolia Cardona, Katy Bell, Justin Clark, Mary-Louise McLaws, and Paul Glasziou. Estimating the extent of asymptomatic COVID-19 and its potential for community transmission: Systematic review and meta-analysis.Journal of the Association of Medical Microbiology and Infectious Disease Canada, 5(4):223–234, 2020. doi: 10.3138/ jammi-2020-0030

    2020

  33. [33]

    The incubation period during the pandemic of COVID- 19: A systematic review and meta-analysis.Systematic Reviews, 10(1):101, 2021

    Wafa Dhouib, Jihen Maatoug, Imen Ayouni, Nawel Zammit, Rim Ghammem, Sihem Ben Fredj, and Hassen Ghannem. The incubation period during the pandemic of COVID- 19: A systematic review and meta-analysis.Systematic Reviews, 10(1):101, 2021. doi: 10.1186/s13643-021-01648-y

    work page doi:10.1186/s13643-021-01648-y 2021

  34. [34]

    Mohr, Sebastian Contreras, Philipp D¨ onges, Emil N

    Jonas Dehning, Sebastian B. Mohr, Sebastian Contreras, Philipp D¨ onges, Emil N. Iftekhar, Oliver Schulz, Philip Bechtle, and Viola Priesemann. Impact of the Euro 2020 cham- pionship on the spread of COVID-19.Nature Communications, 14(1):122, 2023. doi: 10.1038/s41467-022-35512-x

    work page doi:10.1038/s41467-022-35512-x 2020

  35. [35]

    C. J. F. ter Braak and J. A. Vrugt. Differential evolution markov chain with snooker updater and fewer chains.Statistics and Computing, 18(4):435–446, 2008. doi: 10.1007/s11222-008-9104-9

    work page doi:10.1007/s11222-008-9104-9 2008

  36. [36]

    Fonnesbeck and Maxim Kochurov and Ravin Kumar and Junpeng Lao and Christian C

    Oriol Abril-Pla, Virgile Andreani, Colin Carroll, Larry Dong, Christopher J. Fonnesbeck, Maxim Kochurov, Ravin Kumar, Junpeng Lao, Christian C. Luhmann, Osvaldo A. Martin, Michael Osthege, Ricardo Vieira, Thomas Wiecki, and Robert Zinkov. PyMC: A modern and comprehensive probabilistic programming framework in Python.PeerJ Computer Science, 9 (e1516), 2023...

    work page doi:10.7717/peerj-cs.1516 2023

  37. [37]

    Vo, H.X., Durlofsky, L.J.,

    Aki Vehtari, Andrew Gelman, Daniel Simpson, Bob Carpenter, and Paul-Christian B¨ urkner. Rank-normalization, folding, and localization: An improved bR for assessing convergence of MCMC (with discussion).Bayesian Analysis, 16(2):667–718, 2021. doi: 10.1214/20-BA1221

    work page doi:10.1214/20-ba1221 2021

  38. [38]

    Jan-Matthis Lueckmann, Giacomo Bassetto, Theofanis Karaletsos, and Jakob H. Macke. Likelihood-free inference with emulator networks, 2019

    2019

  39. [39]

    Masked autoregressive flow for density estimation, 2018

    George Papamakarios, Theo Pavlakou, and Iain Murray. Masked autoregressive flow for density estimation, 2018. 24

    2018

  40. [40]

    Lecun, L

    Y. Lecun, L. Bottou, Y. Bengio, and P. Haffner. Gradient-based learning applied to document recognition.Proceedings of the IEEE, 86(11):2278–2324, 1998. doi: 10.1109/5.726791

    work page doi:10.1109/5.726791 1998

  41. [41]

    Greenberg, Marcel Nonnenmacher, and Jakob H

    David S. Greenberg, Marcel Nonnenmacher, and Jakob H. Macke. Automatic posterior transformation for likelihood-free inference, 2019

    2019

  42. [42]

    Gonçalves and David S

    Alvaro Tejero-Cantero, Jan Boelts, Michael Deistler, Jan-Matthis Lueckmann, Conor Durkan, Pedro J. Gon¸ calves, David S. Greenberg, and Jakob H. Macke. sbi: A toolkit for simulation- based inference.Journal of Open Source Software, 5(52):2505, 2020. doi: 10.21105/joss.02505

    work page doi:10.21105/joss.02505 2020

  43. [43]

    Vollert, Christopher Drovandi, Cailan Jeynes-Smith, Luz V

    Sarah A. Vollert, Christopher Drovandi, Cailan Jeynes-Smith, Luz V. Pascal, and Matthew P. Adams. Beyond data: Leveraging non-empirical information and expert knowledge in bayesian model calibration, 2025

    2025

  44. [44]

    Henrik Zunker, Philipp D¨ onges, Patrick Lenz, Seba Contreras, and Martin J. K¨ uhn. Risk- mediated dynamic regulation of effective contacts de-synchronizes outbreaks in metapopulation epidemic models.Chaos, Solitons & Fractals, 199:116782, 2025. ISSN 0960-0779. doi: 10.1016/ j.chaos.2025.116782

    arXiv 2025

  45. [45]

    Fast ϵ-free inference of simulation models with bayesian conditional density estimation, 2018

    George Papamakarios and Iain Murray. Fast ϵ-free inference of simulation models with bayesian conditional density estimation, 2018

    2018

  46. [46]

    On large-batch training for deep learning: Generalization gap and sharp minima

    Nitish Shirish Keskar, Dheevatsa Mudigere, Jorge Nocedal, Mikhail Smelyanskiy, and Ping Tak Peter Tang. On large-batch training for deep learning: Generalization gap and sharp minima. InInternational Conference on Learning Representations, 2017. URL https:// openreview.net/forum?id=H1oyRlYgg

    2017

  47. [47]

    Revisiting small batch training for deep neural networks,

    Dominic Masters and Carlo Luschi. Revisiting small batch training for deep neural networks,

  48. [48]

    URLhttps://arxiv.org/abs/1804.07612

    Pith/arXiv arXiv

  49. [49]

    J¨ ulich Supercomputing Centre. JURECA: Data Centric and Booster Modules implementing the Modular Supercomputing Architecture at J¨ ulich Supercomputing Centre.Journal of large-scale research facilities, 7(A182), 2021. doi: 10.17815/jlsrf-7-182. URL http://dx.doi.org/10. 17815/jlsrf-7-182

    work page doi:10.17815/jlsrf-7-182 2021

  50. [50]

    Made: Masked autoencoder for distribution estimation, 2015

    Mathieu Germain, Karol Gregor, Iain Murray, and Hugo Larochelle. Made: Masked autoencoder for distribution estimation, 2015. arXiv:1502.03509. 25 Supplementary information S1 Mean Autoregressive Flows Generally, normalizing flows learn the distribution of the data X p X(x; ϕ) through a mapping fϕ : Rn →R n between the latent variables Z and the data X, wh...

    Pith/arXiv arXiv 2015