pith. machine review for the scientific record. sign in

arxiv: 2605.09870 · v1 · submitted 2026-05-11 · 💻 cs.LG · cs.AI

Recognition: no theorem link

Intervention-Based Time Series Causal Discovery via Simulator-Generated Interventional Distributions

Tsuyoshi Okita
Authors on Pith no claims yet

Pith reviewed 2026-05-12 04:56 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords time series causal discoverystructural VARflow matchinginterventional distributionssimulator-based inferencecausal identifiabilityconfounding removal
0
0 comments X

The pith

A physics-based simulator generates interventional data that makes the full structural VAR in time series identifiable.

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

The paper proposes using a physics-based simulator to generate interventional distributions by clamping variables inside it, which severs confounding paths by construction. Conditional Flow Matching then learns the nonlinear interventional conditionals directly from this data. The authors prove that the complete structural VAR becomes identifiable when the simulator can clamp a sufficient set of variables, and they derive an error bound that breaks down into Monte Carlo sampling, simulator fidelity, and flow matching components. They also establish a sign-flip corollary: below a fidelity threshold the estimated causal effect reverses sign. This matters because observational time series data alone frequently produces reversed causal signs due to hidden confounders.

Core claim

SVAR-FM treats the simulator as a mechanical realization of Pearl's do-operator via variable clamping to produce interventional data by construction, then uses Conditional Flow Matching to recover the structural vector autoregression. Under a coverage condition on the clampable variables the full model is identifiable, with an end-to-end error bound separating Monte Carlo, simulator fidelity, and Flow Matching terms. A corollary predicts and a laser-physics case study confirms that causal-effect signs reverse when simulator accuracy falls below a threshold.

What carries the argument

Simulator clamping as a direct mechanical do-operator, paired with Conditional Flow Matching to learn interventional conditionals from the generated distributions.

If this is right

  • The full causal structure of a time series becomes recoverable even when observational methods are misled by confounding.
  • Total estimation error decomposes into separate Monte Carlo, simulator fidelity, and learning terms that can be bounded and reduced independently.
  • Causal effect signs flip when simulator fidelity drops below a critical threshold.
  • Correct causal signs are recovered across four scientific domains and a laser-physics case study where baselines produce reversed estimates.

Where Pith is reading between the lines

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

  • Improving simulator fidelity may be more effective for causal discovery than collecting more observational data.
  • The sign-flip prediction could serve as a practical diagnostic for whether a simulator is accurate enough to trust for causal claims.
  • The coverage condition on clampable variables suggests testing the method on simulators with only partial intervention access to see where identifiability breaks.
  • The error-bound decomposition could guide allocation of computational resources between more simulator runs and better flow-matching models.

Load-bearing premise

Clamping a variable inside the simulator must physically cut all confounding paths exactly as a real-world intervention would.

What would settle it

In a controlled physical experiment, if the estimated causal sign does not reverse precisely when simulator accuracy is lowered below the predicted threshold, or if the sign remains wrong even with high-accuracy simulation that matches real data, the identifiability claim fails.

Figures

Figures reproduced from arXiv: 2605.09870 by Tsuyoshi Okita.

Figure 1
Figure 1. Figure 1: Architecture of SVAR-FM. The inputs are the observed time series [PITH_FULL_IMAGE:figures/full_fig_p015_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: CausalSim-Macro: method comparison (50 seeds). SVAR-FM (orange) achieves the smallest [PITH_FULL_IMAGE:figures/full_fig_p022_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: CausalSim-Diabetes: method comparison (20 seeds). SVAR-FM (orange) achieves the smallest [PITH_FULL_IMAGE:figures/full_fig_p023_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: CausalSim-Cosmic: results from the cosmic ray shower experiment (KASCADE real data, 3.34 [PITH_FULL_IMAGE:figures/full_fig_p025_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Method comparison for HHG causal effect estimation (real Octopus SIC-ADSIC data). [PITH_FULL_IMAGE:figures/full_fig_p029_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Causal graphs identified by each method in HHG. [PITH_FULL_IMAGE:figures/full_fig_p030_6.png] view at source ↗
read the original abstract

We propose SVAR-FM (Structural VAR with Flow Matching), a framework for time series causal discovery that treats a physics-based simulator as a mechanical realization of Pearl's do operator. Clamping a variable inside the simulator physically severs confounding paths, producing interventional data by construction. Conditional Flow Matching then learns the nonlinear interventional conditionals. Theoretically, we prove that the full structural VAR becomes identifiable under a coverage condition on the simulator-clampable variables, and derive an end-to-end error bound that decomposes into Monte Carlo, simulator fidelity, and Flow Matching terms. A sign-flip corollary predicts that when simulator accuracy falls below a threshold, the estimated causal effect reverses sign. Empirically, a benchmark across four scientific domains confirms that SVAR-FM recovers the correct causal sign where observational methods produce sign-reversed estimates due to confounding. A case study in ultrafast laser physics verifies the sign-flip prediction by physically varying the accuracy level of a first-principles quantum solver: the low-accuracy setting reverses the causal sign, while the high-accuracy setting recovers the correct direction (R-squared = 0.983, zero bias).

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 proposes SVAR-FM, a framework for time series causal discovery that treats a physics-based simulator as a direct mechanical realization of Pearl's do-operator. By clamping variables inside the simulator, interventional distributions are generated by construction. Conditional flow matching is then applied to learn the nonlinear interventional conditionals of a structural VAR. The authors prove identifiability of the full structural VAR under a coverage condition on the simulator-clampable variables, derive an end-to-end error bound decomposing into Monte Carlo, simulator fidelity, and flow matching terms, and state a sign-flip corollary. Empirical support includes benchmarks across four scientific domains (recovering correct causal signs where observational methods fail due to confounding) and a case study in ultrafast laser physics that physically varies simulator accuracy to confirm the sign-flip prediction (R²=0.983, zero bias).

Significance. If the derivations hold, the work is significant because it supplies a principled route to interventional data for causal discovery in domains equipped with high-fidelity simulators, directly addressing confounding that defeats purely observational methods. The explicit error decomposition and the falsifiable sign-flip corollary supply practical guidance on required simulator accuracy. The empirical verification across multiple domains plus the physical experiment in laser physics strengthens the contribution. The paper earns credit for stating a clear identifiability result, an end-to-end error bound, and a testable prediction that is then checked experimentally.

major comments (2)
  1. [Theoretical section (identifiability theorem)] The coverage condition on clampable variables is load-bearing for the identifiability claim; the precise statement of this condition (including the minimal number and type of variables that must be clampable) should be given explicitly together with the theorem, and the interaction between non-zero simulator fidelity and the coverage requirement should be clarified.
  2. [Error-bound derivation and corollary] The sign-flip corollary is derived from the error bound; the manuscript should show the explicit threshold on simulator fidelity at which the estimated causal effect changes sign, and confirm that this threshold is independent of the particular flow-matching implementation.
minor comments (2)
  1. [Abstract] The abstract refers to benchmarks in four scientific domains without naming them; listing the domains in the abstract or introduction would improve immediate readability.
  2. [Methods] Notation for the structural VAR and the flow-matching objective is introduced late; moving the core definitions to the beginning of the methods section would aid readers.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive evaluation and the recommendation for minor revision. The comments highlight opportunities to strengthen the presentation of the theoretical results. We address each point below and will revise the manuscript accordingly to make the coverage condition and the sign-flip threshold fully explicit.

read point-by-point responses

  1. Referee: [Theoretical section (identifiability theorem)] The coverage condition on clampable variables is load-bearing for the identifiability claim; the precise statement of this condition (including the minimal number and type of variables that must be clampable) should be given explicitly together with the theorem, and the interaction between non-zero simulator fidelity and the coverage requirement should be clarified.

    Authors: We agree that the coverage condition should be stated with greater precision. In the revised manuscript we will restate the identifiability theorem immediately after its proof and specify that the coverage condition requires every variable appearing in the structural VAR to be clampable inside the simulator; this is the minimal requirement that severs all confounding paths and renders the full set of interventional conditionals identifiable. Regarding the interaction with simulator fidelity, the end-to-end error bound already isolates the fidelity term from the coverage assumption: coverage guarantees identifiability in the zero-fidelity-error limit, while the bound quantifies the additional deviation introduced by non-zero simulator error. We will add a short remark after the theorem making this separation explicit. revision: yes

  2. Referee: [Error-bound derivation and corollary] The sign-flip corollary is derived from the error bound; the manuscript should show the explicit threshold on simulator fidelity at which the estimated causal effect changes sign, and confirm that this threshold is independent of the particular flow-matching implementation.

    Authors: The sign-flip corollary is obtained by requiring the total error (Monte Carlo + fidelity + flow-matching) to exceed the magnitude of the true causal effect. In the revision we will insert the explicit threshold: the estimated effect reverses sign whenever simulator fidelity error exceeds |true effect| divided by the constant factor arising from the bound (specifically, the sum of the Lipschitz constants of the structural functions). Because the error bound is derived from general approximation and sampling arguments that do not depend on the internal details of the flow-matching procedure, the resulting threshold is independent of any particular flow-matching implementation. We will add a sentence immediately after the corollary stating this independence. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected in derivation chain

full rationale

The paper grounds its central mechanism in an external physics-based simulator that is posited to realize Pearl's do-operator via variable clamping, producing interventional distributions by construction as an independent modeling assumption rather than an internal derivation. The identifiability theorem under the coverage condition on clampable variables and the end-to-end error bound (decomposing into Monte Carlo, simulator fidelity, and Flow Matching terms) reference quantities that are not tautological with the target causal estimates. The sign-flip corollary is derived from the error bound and then subjected to direct physical verification by varying simulator accuracy in the laser-physics case study. No load-bearing step reduces by the paper's own equations to a fitted parameter renamed as a prediction, a self-citation chain, or a self-definitional construct; the framework remains self-contained against external benchmarks and empirical checks across domains.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The identifiability result and error bound rest on the simulator faithfully implementing interventions and on a coverage condition whose precise statement is not expanded in the abstract.

axioms (2)
  • domain assumption Coverage condition on the simulator-clampable variables
    Invoked to guarantee that the full structural VAR is identifiable.
  • domain assumption Simulator clamping physically severs confounding paths
    Required for the claim that interventional data is produced by construction.

pith-pipeline@v0.9.0 · 5491 in / 1337 out tokens · 53017 ms · 2026-05-12T04:56:37.052359+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

258 extracted references · 258 canonical work pages · 10 internal anchors

  1. [1]

    Granger, C. W. J. , title =. Econometrica , year =

    work page

  2. [2]

    , title =

    Sims, Christopher A. , title =. Econometrica , year =

    work page

  3. [3]

    Structural Vector Autoregressive Analysis , publisher =

    Kilian, Lutz and L. Structural Vector Autoregressive Analysis , publisher =. 2017 , address =

    work page 2017

  4. [4]

    and Quah, Danny , title =

    Blanchard, Olivier J. and Quah, Danny , title =. American Economic Review , year =

    work page

  5. [5]

    Journal of Monetary Economics , year =

    Uhlig, Harald , title =. Journal of Monetary Economics , year =

    work page

  6. [6]

    Estimation of a Structural Vector Autoregression Model Using Non-

    Hyv. Estimation of a Structural Vector Autoregression Model Using Non-. Journal of Machine Learning Research , year =

    work page

  7. [7]

    , title =

    Hamilton, James D. , title =. 1994 , address =

    work page 1994

  8. [8]

    and Janzing, Dominik and Mooij, Joris M

    Hoyer, Patrik O. and Janzing, Dominik and Mooij, Joris M. and Peters, Jonas and Sch. Nonlinear Causal Discovery with Additive Noise Models , booktitle =. 2009 , pages =

    work page 2009

  9. [9]

    , title =

    Zheng, Xun and Aragam, Bryon and Ravikumar, Pradeep and Xing, Eric P. , title =. Advances in Neural Information Processing Systems 31 (NeurIPS 2018) , year =

    work page 2018

  10. [10]

    Differentiable Causal Discovery from Interventional Data , booktitle =

    Brouillard, Philippe and Lachapelle, S. Differentiable Causal Discovery from Interventional Data , booktitle =. 2020 , pages =

    work page 2020

  11. [11]

    2009 , address =

    Pearl, Judea , title =. 2009 , address =

    work page 2009

  12. [12]

    Proceedings of the 21st Conference on Uncertainty in Artificial Intelligence (UAI 2005) , year =

    Eberhardt, Frederick and Glymour, Clark and Scheines, Richard , title =. Proceedings of the 21st Conference on Uncertainty in Artificial Intelligence (UAI 2005) , year =

    work page 2005

  13. [13]

    Philosophy of Science , year =

    Eberhardt, Frederick and Glymour, Clark and Scheines, Richard , title =. Philosophy of Science , year =

    work page

  14. [14]

    Characterization and Greedy Learning of Interventional

    Hauser, Alain and B. Characterization and Greedy Learning of Interventional. Journal of Machine Learning Research , year =

    work page

  15. [15]

    Lipman, Yaron and Chen, Ricky T. Q. and Ben-Hamu, Heli and Nickel, Maximilian and Le, Matt , title =. The Eleventh International Conference on Learning Representations (ICLR 2023) , year =

    work page 2023

  16. [16]

    and Vanden-Eijnden, Eric , title =

    Albergo, Michael S. and Vanden-Eijnden, Eric , title =. The Eleventh International Conference on Learning Representations (ICLR 2023) , year =

    work page 2023

  17. [17]

    Transactions on Machine Learning Research , year =

    Tong, Alexander and Malkin, Nikolay and Huguet, Guillaume and Zhang, Yanlei and Rector-Brooks, Jarrid and Fatras, Kilian and Wolf, Guy and Bengio, Yoshua , title =. Transactions on Machine Learning Research , year =

    work page

  18. [18]

    Chen, Ricky T. Q. and Rubanova, Yulia and Bettencourt, Jesse and Duvenaud, David , title =. Advances in Neural Information Processing Systems 31 (NeurIPS 2018) , year =

    work page 2018

  19. [19]

    van der Vaart, A. W. , title =. 2000 , address =

    work page 2000

  20. [20]

    2006 , address =

    Wasserman, Larry , title =. 2006 , address =

    work page 2006

  21. [21]

    Journal of the American Statistical Association , year =

    Hoeffding, Wassily , title =. Journal of the American Statistical Association , year =

    work page

  22. [22]

    , title =

    Valiant, Leslie G. , title =. Communications of the ACM , year =

    work page

  23. [23]

    and Vazirani, Umesh V

    Kearns, Michael J. and Vazirani, Umesh V. , title =. 1994 , address =

    work page 1994

  24. [24]

    Saha, Bhaskar and Goebel, Kai , title =

    work page

  25. [25]

    and Sohn, Jerry J

    Bloom, Ira and Cole, Brian W. and Sohn, Jerry J. and Jones, Steven A. and Polzin, Edward G. and Battaglia, Vincent S. and Henriksen, Gary L. and Motloch, Chester and Richardson, Roberta and Unkelhaeuser, Tony and Ingersoll, David and Case, Howard L. , title =. Journal of Power Sources , year =

    work page

  26. [26]

    and Attia, Peter M

    Severson, Kristen A. and Attia, Peter M. and Jin, Norman and Perkins, Nicholas and Jiang, Benben and Yang, Zi and Chen, Michael H. and Aykol, Muratahan and Herring, Patrick K. and Fraggedakis, Dimitrios and Bazant, Martin Z. and Harris, Stephen J. and Chueh, William C. and Braatz, Richard D. , title =. Nature Energy , year =

    work page

  27. [27]

    Journal of Diabetes Science and Technology , year =

    Man, Chiara Dalla and Micheletto, Francesco and Lv, Dayu and Breton, Marc and Kovatchev, Boris and Cobelli, Claudio , title =. Journal of Diabetes Science and Technology , year =

    work page

  28. [28]

    and Breton, Marc and Dalla Man, Chiara and Cobelli, Claudio , title =

    Kovatchev, Boris P. and Breton, Marc and Dalla Man, Chiara and Cobelli, Claudio , title =. Journal of Diabetes Science and Technology , year =

    work page

  29. [29]

    Journal of Diabetes Science and Technology , year =

    Cobelli, Claudio and Kovatchev, Boris , title =. Journal of Diabetes Science and Technology , year =

    work page

  30. [30]

    Xie, Jinyu , title =

    work page

  31. [31]

    Scientific Data , year =

    Zhao, Qiuling and Zhu, Jingyi and Fang, Qingqing and Pan, Yufan and Wang, Yanling and Shao, Jing and Xie, Wenhui , title =. Scientific Data , year =

    work page

  32. [32]

    European Physical Journal C , year =

    Haungs, Andreas and others , title =. European Physical Journal C , year =

    work page

  33. [33]

    , title =

    Matthews, J. , title =. Astroparticle Physics , year =

    work page

  34. [34]

    and others , title =

    Antoni, T. and others , title =. Nuclear Instruments and Methods in Physics Research A , year =

    work page

  35. [35]

    and L'Huillier, Anne and Corkum, Paul B

    Lewenstein, Maciej and Balcou, Philippe and Ivanov, Misha Yu. and L'Huillier, Anne and Corkum, Paul B. , title =. Physical Review A , year =

    work page

  36. [36]

    , title =

    Corkum, Paul B. , title =. Physical Review Letters , year =

    work page

  37. [37]

    Tancogne-Dejean, Nicolas and Oliveira, Micael J. T. and Andrade, Xavier and Appel, Heiko and Borca, Carlos H. and Le Breton, Guillaume and Buchholz, Florian and Castro, Alberto and Corni, Stefano and Correa, Alfredo A. and De Giovannini, Umberto and Delgado, Alain and Eich, F. G. and Flick, Johannes and Gil, Gabriel and Gomez, Alejandro and Helbig, Nicole...

    work page 2020

  38. [38]

    Science Advances , year =

    Runge, Jakob and Nowack, Peer and Kretschmer, Marlene and Flaxman, Seth and Sejdinovic, Dino , title =. Science Advances , year =

    work page

  39. [39]

    Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence (UAI 2020) , year =

    Runge, Jakob , title =. Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence (UAI 2020) , year =

    work page 2020

  40. [40]

    Advances in Neural Information Processing Systems 33 (NeurIPS 2020) , year =

    Gerhardus, Andreas and Runge, Jakob , title =. Advances in Neural Information Processing Systems 33 (NeurIPS 2020) , year =

    work page 2020

  41. [41]

    , title =

    Tank, Alex and Covert, Ian and Foti, Nicholas and Shojaie, Ali and Fox, Emily B. , title =. IEEE Transactions on Pattern Analysis and Machine Intelligence , year =

    work page

  42. [42]

    Proceedings of the National Academy of Sciences , year =

    Cranmer, Kyle and Brehmer, Johann and Louppe, Gilles , title =. Proceedings of the National Academy of Sciences , year =

    work page

  43. [43]

    Advances in Neural Information Processing Systems 33 (NeurIPS 2020) , year =

    Pawlowski, Nick and Castro, Daniel Coelho and Glocker, Ben , title =. Advances in Neural Information Processing Systems 33 (NeurIPS 2020) , year =

    work page 2020

  44. [44]

    Causal Autoregressive Flows , booktitle =

    Khemakhem, Ilyes and Monti, Ricardo Pio and Leech, Robert and Hyv. Causal Autoregressive Flows , booktitle =. 2021 , pages =

    work page 2021

  45. [45]

    Causal Normalizing Flows: From Theory to Practice , booktitle =

    Javaloy, Adri. Causal Normalizing Flows: From Theory to Practice , booktitle =

    work page

  46. [46]

    Modeling Causal Mechanisms with Diffusion Models for Interventional and Counterfactual Queries , journal =

    Chao, Patrick and Bl. Modeling Causal Mechanisms with Diffusion Models for Interventional and Counterfactual Queries , journal =

    work page

  47. [49]

    Advances in Neural Information Processing Systems 38 (NeurIPS 2025) , year =

    Almod. Advances in Neural Information Processing Systems 38 (NeurIPS 2025) , year =

    work page 2025

  48. [51]

    Transactions on Machine Learning Research , year =

    Komanduri, Aneesh and Wu, Xintao and Wu, Yongkai and Chen, Feng , title =. Transactions on Machine Learning Research , year =

    work page

  49. [52]

    Proceedings of the AAAI Conference on Artificial Intelligence , year =

    Le, Minh Khoa and Do, Kien and Tran, Truyen , title =. Proceedings of the AAAI Conference on Artificial Intelligence , year =

    work page

  50. [53]

    , title =

    Sanchez, Pedro and Tsaftaris, Sotirios A. , title =. Proceedings of the First Conference on Causal Learning and Reasoning (CLEaR) , series =. 2022 , pages =

    work page 2022

  51. [54]

    Nature Communications , year =

    Park, Sung Hoon and Ha, Seokmin and Kim, Jae Kyoung , title =. Nature Communications , year =

    work page

  52. [55]

    Proceedings of the National Academy of Sciences , year =

    Brehmer, Johann and Louppe, Gilles and Pavez, Juan and Cranmer, Kyle , title =. Proceedings of the National Academy of Sciences , year =

    work page

  53. [56]

    Proceedings of the 41st Conference on Uncertainty in Artificial Intelligence (UAI 2025) , series =

    Misiakos, Panagiotis and P. Proceedings of the 41st Conference on Uncertainty in Artificial Intelligence (UAI 2025) , series =. 2025 , pages =

    work page 2025

  54. [57]

    2025 , type =

    Brignone, Dario and Piffer, Michele , title =. 2025 , type =

    work page 2025

  55. [58]

    Causal Digital Twins for Cyber-Physical Security in Water Systems: A Framework for Robust Anomaly Detection , journal =

    Homaei, Mohammad Hadi and Tarif, Mohammed and Rodr. Causal Digital Twins for Cyber-Physical Security in Water Systems: A Framework for Robust Anomaly Detection , journal =. 2025 , volume =

    work page 2025

  56. [59]

    Proceedings of the Third Conference on Causal Learning and Reasoning (CLEaR) , series =

    Rohbeck, Martin and Clarke, Brendan and Mikulik, Kris and Pettet, Andrew and Stegle, Oliver and Ueltzhoeffer, Kai , title =. Proceedings of the Third Conference on Causal Learning and Reasoning (CLEaR) , series =. 2024 , pages =

    work page 2024

  57. [60]

    , title =

    Raissi, Maziar and Perdikaris, Paris and Karniadakis, George E. , title =. Journal of Computational Physics , year =

    work page

  58. [61]

    Journal of the Royal Statistical Society: Series B , year =

    Benjamini, Yoav and Hochberg, Yosef , title =. Journal of the Royal Statistical Society: Series B , year =

    work page

  59. [62]

    The Twelfth International Conference on Learning Representations (ICLR 2024) , year =

    Cheng, Hao and others , title =. The Twelfth International Conference on Learning Representations (ICLR 2024) , year =

    work page 2024

  60. [63]

    Advances in Neural Information Processing Systems 38 (NeurIPS 2025) , year =

    Herdeanu, Bohdan and others , title =. Advances in Neural Information Processing Systems 38 (NeurIPS 2025) , year =

    work page 2025

  61. [64]

    The Ninth International Conference on Learning Representations (ICLR 2021) , year =

    Li, Zongyi and Kovachki, Nikola and Azizzadenesheli, Kamyar and Liu, Burigede and Bhattacharya, Kaushik and Stuart, Andrew and Anandkumar, Anima , title =. The Ninth International Conference on Learning Representations (ICLR 2021) , year =

    work page 2021

  62. [65]

    Oxford Statistical Science Series , year =

    Dahlhaus, Rainer and Eichler, Michael , title =. Oxford Statistical Science Series , year =

    work page

  63. [66]

    Probability Theory and Related Fields , year =

    Eichler, Michael , title =. Probability Theory and Related Fields , year =

    work page

  64. [67]

    Nature Reviews Earth & Environment , year =

    Runge, Jakob and Gerhardus, Andreas and Varando, Gherardo and Eyring, Veronika and Camps-Valls, Gustau , title =. Nature Reviews Earth & Environment , year =

    work page

  65. [68]

    Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics (AISTATS 2020) , year =

    Pamfil, Roxana and Sriwattanaworachai, Nisara and Desai, Shaan and Pilgerstorfer, Philip and Georgatzis, Konstantinos and Beaumont, Paul and Aragam, Bryon , title =. Proceedings of the 23rd International Conference on Artificial Intelligence and Statistics (AISTATS 2020) , year =

    work page 2020

  66. [69]

    The Eleventh International Conference on Learning Representations (ICLR 2023) , year =

    Gong, Wenbo and Jennings, Joel and Zhang, Cheng and Pawlowski, Nick , title =. The Eleventh International Conference on Learning Representations (ICLR 2023) , year =

    work page 2023

  67. [70]

    The Eleventh International Conference on Learning Representations (ICLR 2023) , year =

    Cheng, Yuxiao and Yang, Runzhao and Xiao, Tingxiong and Li, Zongren and Suo, Jinli and He, Kunlun and Dai, Qionghai , title =. The Eleventh International Conference on Learning Representations (ICLR 2023) , year =

    work page 2023

  68. [71]

    Amortized Causal Discovery: Learning to Infer Causal Graphs from Time-Series Data , booktitle =

    L. Amortized Causal Discovery: Learning to Infer Causal Graphs from Time-Series Data , booktitle =. 2022 , series =

    work page 2022

  69. [72]

    and Devijver, Emilie and Gaussier, Eric , title =

    Assaad, Charles K. and Devijver, Emilie and Gaussier, Eric , title =. Journal of Artificial Intelligence Research , year =

    work page

  70. [73]

    Proceedings of Machine Learning Research (AISTATS 2023) , year =

    Sun, Xiangyu and Schulte, Oliver and Liu, Guiliang and Poupart, Pascal , title =. Proceedings of Machine Learning Research (AISTATS 2023) , year =

    work page 2023

  71. [74]

    and Uhler, Caroline , title =

    Wang, Yuhao and Solus, Liam and Yang, Karren D. and Uhler, Caroline , title =. Advances in Neural Information Processing Systems 30 (NeurIPS 2017) , year =

    work page 2017

  72. [75]

    Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence (UAI 2020) , year =

    Squires, Chandler and Wang, Yuhao and Uhler, Caroline , title =. Proceedings of the 36th Conference on Uncertainty in Artificial Intelligence (UAI 2020) , year =

    work page 2020

  73. [76]

    and Magliacane, Sara and Claassen, Tom , title =

    Mooij, Joris M. and Magliacane, Sara and Claassen, Tom , title =. Journal of Machine Learning Research , year =

    work page

  74. [77]

    The Tenth International Conference on Learning Representations (ICLR 2022) , year =

    Lippe, Phillip and Cohen, Taco and Gavves, Efstratios , title =. The Tenth International Conference on Learning Representations (ICLR 2022) , year =

    work page 2022

  75. [78]

    Active Bayesian Causal Inference , booktitle =

    Toth, Christian and Lorch, Lars and Knoll, Christian and Krause, Andreas and Pernkopf, Franz and Peharz, Robert and von K. Active Bayesian Causal Inference , booktitle =. 2022 , pages =

    work page 2022

  76. [79]

    , title =

    Entner, Doris and Hoyer, Patrik O. , title =. Proceedings of the 5th European Workshop on Probabilistic Graphical Models (PGM 2010) , year =

    work page 2010

  77. [80]

    Proceedings of 2018 ACM SIGKDD Workshop on Causal Discovery , year =

    Malinsky, Daniel and Spirtes, Peter , title =. Proceedings of 2018 ACM SIGKDD Workshop on Causal Discovery , year =

    work page 2018

  78. [81]

    On the Identifiability of the Post-Nonlinear Causal Model , journal =

    Zhang, Kun and Hyv. On the Identifiability of the Post-Nonlinear Causal Model , journal =. 2009 , pages =

    work page 2009

  79. [82]

    Score Matching Enables Causal Discovery of Nonlinear Additive Noise Models , booktitle =

    Rolland, Paul and Cevher, Volkan and Kleindessner, Matth. Score Matching Enables Causal Discovery of Nonlinear Additive Noise Models , booktitle =. 2022 , series =

    work page 2022

  80. [83]

    Proceedings of the Second Conference on Causal Learning and Reasoning (CLeaR 2023) , year =

    Montagna, Francesco and Noceti, Nicoletta and Rosasco, Lorenzo and Zhang, Kun and Locatello, Francesco , title =. Proceedings of the Second Conference on Causal Learning and Reasoning (CLeaR 2023) , year =

    work page 2023

Showing first 80 references.