pith. machine review for the scientific record. sign in

arxiv: 2605.07267 · v1 · submitted 2026-05-08 · 💻 cs.LG

Recognition: 2 theorem links

· Lean Theorem

PerCaM-Health: Personalized Dynamic Causal Graphs for Healthcare Reasoning

Amir M. Rahmani, Di Huang, Elahe Khatibi, Hung Cao, Ramesh Jain, Saba A. Farahani, Ziyu Wang

Authors on Pith no claims yet

Pith reviewed 2026-05-11 01:37 UTC · model grok-4.3

classification 💻 cs.LG
keywords personalized causal discoverydynamic causal graphslongitudinal health datacounterfactual reasoningtemporal adaptationintervention analysishealthcare modeling
0
0 comments X

The pith

PerCaM-Health adapts a population causal graph with each patient's own data to produce reliable time-varying structures for intervention reasoning.

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

The paper seeks to close the gap between stable but generic population causal models and unreliable per-patient models by building a framework that begins with a knowledge-guided temporal graph and then updates it conservatively for individual patients. This matters because effective healthcare decisions depend on understanding how variables influence one specific person over time, yet short and irregular patient records make direct discovery unreliable. The method produces sequences of evolving graphs paired with structural equations that support counterfactual questions about short-term outcome changes under hypothetical interventions. Experiments on a semi-synthetic benchmark show gains in graph recovery, edge tracking, and intervention accuracy over cohort-level and per-patient baselines.

Core claim

PerCaM-Health learns a knowledge-guided population temporal graph, then conservatively adapts and evolves it using patient-specific temporal evidence and rolling-window updates, producing interpretable and auditable graph sequences. By coupling these graphs with temporal structural equations, the framework enables patient-level counterfactual queries, such as estimating short-horizon outcome changes under hypothetical behavioral interventions.

What carries the argument

Conservative adaptation and rolling-window evolution of a knowledge-guided population temporal graph using short patient-specific evidence, which generates dynamic personalized causal structures for counterfactual reasoning.

If this is right

  • Better recovery of true causal structures on semi-synthetic dynamic health benchmarks.
  • Improved tracking of how causal edges change over time for individual patients.
  • Higher accuracy in identifying the direction and effect of behavioral interventions.
  • More reliable patient-level causal reasoning compared to purely cohort or purely individual methods.

Where Pith is reading between the lines

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

  • The rolling-window mechanism could support ongoing updates in live clinical systems where new measurements arrive continuously.
  • Similar adaptation logic might apply to other domains that combine population knowledge with sparse individual trajectories, such as personalized education or financial forecasting.
  • Coupling the graphs with structural equations creates a pathway for simulation of treatment sequences that could be tested against held-out patient outcomes.

Load-bearing premise

A knowledge-guided population temporal graph can be conservatively adapted and evolved using short, noisy, irregular, non-stationary patient-specific evidence without introducing errors that undermine the counterfactual queries.

What would settle it

A controlled test on real longitudinal patient records where known interventions occur and the model's predicted counterfactual outcome changes diverge substantially from the observed changes.

Figures

Figures reproduced from arXiv: 2605.07267 by Amir M. Rahmani, Di Huang, Elahe Khatibi, Hung Cao, Ramesh Jain, Saba A. Farahani, Ziyu Wang.

Figure 1
Figure 1. Figure 1: Motivating example. A cohort-level causal prior captures stable population relationships, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of PERCAM-HEALTH: heterogeneous health data are converted into patient-state trajectories, combined with clinical priors to learn and personalize dynamic causal graphs, and used for patient-level counterfactual queries. 3 Method We propose PERCAM-HEALTH, a modular framework for learning patient-specific dynamic causal graphs from longitudinal healthcare data. The method is designed for settings in… view at source ↗
Figure 3
Figure 3. Figure 3: Personalization and dynamic tracking results. Patient-specific adaptation improves precision [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Counterfactual reasoning and ablation results. Personalized dynamic causal reasoning [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
read the original abstract

Personalized healthcare decisions require reasoning about how physiological and behavioral variables influence an individual patient over time. Existing temporal causal discovery methods are poorly matched to this setting: cohort-level models provide stable but non-personalized structures, while per-patient discovery is unreliable because individual trajectories are short, noisy, irregular, and non-stationary. This creates a fundamental gap between population-level causal modeling and the patient-specific, time-varying mechanisms needed for intervention reasoning. We introduce PerCaM-Health, a framework for learning personalized dynamic causal graphs from longitudinal health data. The framework learns a knowledge-guided population temporal graph, then conservatively adapts and evolves it using patient-specific temporal evidence and rolling-window updates, producing interpretable and auditable graph sequences. By coupling these graphs with temporal structural equations, the framework enables patient-level counterfactual queries, such as estimating short-horizon outcome changes under hypothetical behavioral interventions. Experiments on a semi-synthetic dynamic health benchmark show that PerCaM-Health improves graph recovery, dynamic edge tracking, and intervention direction accuracy compared to cohort-level, per-patient, and non-personalized temporal baselines. These results demonstrate that jointly modeling personalization and temporal evolution yields more reliable causal structure and intervention reasoning.

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 introduces PerCaM-Health, a framework for learning personalized dynamic causal graphs from longitudinal health data. It first constructs a knowledge-guided population temporal graph, then conservatively adapts and evolves this graph using patient-specific temporal evidence and rolling-window updates. The resulting graph sequences are coupled with temporal structural equations to support patient-level counterfactual queries, such as short-horizon outcome changes under hypothetical interventions. Experiments on a semi-synthetic dynamic health benchmark are reported to show improvements in graph recovery, dynamic edge tracking, and intervention direction accuracy relative to cohort-level, per-patient, and non-personalized temporal baselines.

Significance. If the conservative adaptation mechanism successfully prevents error propagation from short, noisy, and non-stationary patient trajectories, the framework could meaningfully address the gap between stable population causal models and reliable individualized, time-varying reasoning in healthcare. The semi-synthetic benchmark design is well-suited for isolating dynamic recovery properties that are difficult to evaluate in real data.

major comments (2)
  1. Abstract and Experiments section: The central claim of improved graph recovery, dynamic edge tracking, and intervention direction accuracy is stated without any quantitative metrics, statistical tests, error bars, confidence intervals, or implementation details. This absence prevents verification of whether the semi-synthetic benchmark results actually support the reported gains over the three baseline categories.
  2. Framework description (adaptation and rolling-window updates): The conservative adaptation of the population temporal graph with short, irregular, non-stationary patient evidence is presented as the key enabler of reliable counterfactuals, yet no sensitivity analysis, ablation on noise/missingness levels, or formal bound on error propagation is provided. If the adaptation fails to fully constrain changes under realistic conditions, the claimed improvements in edge direction accuracy and counterfactual reliability would not hold.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive feedback and for recognizing the potential of PerCaM-Health to bridge population-level causal models with personalized, time-varying reasoning. We address each major comment below, indicating the revisions we will make.

read point-by-point responses

  1. Referee: Abstract and Experiments section: The central claim of improved graph recovery, dynamic edge tracking, and intervention direction accuracy is stated without any quantitative metrics, statistical tests, error bars, confidence intervals, or implementation details. This absence prevents verification of whether the semi-synthetic benchmark results actually support the reported gains over the three baseline categories.

    Authors: We agree that the abstract summarizes the improvements qualitatively. The experiments section presents quantitative results via tables and figures comparing graph recovery, dynamic edge tracking, and intervention accuracy against the three baseline categories. To ensure full verifiability, we will revise the abstract to report specific metrics (e.g., relative F1-score gains for graph recovery and accuracy improvements for intervention direction) and expand the experiments section with error bars, confidence intervals, statistical significance tests, and additional implementation details on the benchmark and hyperparameters. revision: yes

  2. Referee: Framework description (adaptation and rolling-window updates): The conservative adaptation of the population temporal graph with short, irregular, non-stationary patient evidence is presented as the key enabler of reliable counterfactuals, yet no sensitivity analysis, ablation on noise/missingness levels, or formal bound on error propagation is provided. If the adaptation fails to fully constrain changes under realistic conditions, the claimed improvements in edge direction accuracy and counterfactual reliability would not hold.

    Authors: The conservative adaptation is intended to restrict updates to patient evidence satisfying strict consistency criteria, thereby limiting propagation of errors from short, noisy trajectories; the semi-synthetic results already reflect performance under irregular and non-stationary conditions. We will add sensitivity analyses across noise and missingness levels plus ablations isolating the adaptation and rolling-window components. A general formal bound on error propagation is difficult to obtain without strong assumptions on non-stationarity, but we will provide a more detailed discussion of the update rules' conservative properties together with the new empirical robustness results. revision: partial

standing simulated objections not resolved
  • A formal mathematical bound on error propagation for the conservative adaptation under arbitrary non-stationarity

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper presents a framework that learns a knowledge-guided population temporal graph and then adapts it via patient-specific evidence and rolling-window updates to produce dynamic graphs for counterfactual queries. No equations or steps in the provided description reduce any claimed prediction or result to a quantity defined by construction from the inputs, fitted parameters, or self-citations. Experimental comparisons to cohort-level, per-patient, and non-personalized baselines are independent of the method's internal definitions, confirming the derivation chain is self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

Review performed on abstract only; full details on parameters, assumptions, and entities are unavailable.

axioms (1)
  • domain assumption Longitudinal health data consists of short, noisy, irregular, and non-stationary trajectories per patient
    Explicitly stated as the reason per-patient discovery is unreliable.
invented entities (1)
  • Personalized dynamic causal graph no independent evidence
    purpose: To represent individual time-varying causal mechanisms
    Core output of the framework introduced to enable patient-level counterfactuals.

pith-pipeline@v0.9.0 · 5526 in / 1116 out tokens · 49093 ms · 2026-05-11T01:37:31.628359+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages · 2 internal anchors

  1. [1]

    Seal: Sensing efficient active learning on wearables through context-awareness

    Hamidreza Alikhani, Ziyu Wang, Anil Kanduri, Pasi Lilieberg, Amir M Rahmani, and Nikil Dutt. Seal: Sensing efficient active learning on wearables through context-awareness. In2024 Design, Automation & Test in Europe Conference & Exhibition (DATE), pages 1–2. IEEE, 2024

    work page 2024

  2. [2]

    Eaˆ 2: Energy efficient adaptive active learning for smart wearables

    Hamidreza Alikhani, Ziyu Wang, Anil Kanduri, Pasi Liljeberg, Amir M Rahmani, and Nikil Dutt. Eaˆ 2: Energy efficient adaptive active learning for smart wearables. InProceedings of the 29th ACM/IEEE international symposium on low power electronics and design, pages 1–6, 2024

    work page 2024

  3. [3]

    Enhancing performance and user engagement in everyday stress monitoring: A context-aware active reinforcement learning approach.arXiv preprint arXiv:2407.08215, 2024

    Seyed Amir Hossein Aqajari, Ziyu Wang, Ali Tazarv, Sina Labbaf, Salar Jafarlou, Brenda Nguyen, Nikil Dutt, Marco Levorato, and Amir M Rahmani. Enhancing performance and user engagement in everyday stress monitoring: A context-aware active reinforcement learning approach.arXiv preprint arXiv:2407.08215, 2024

    work page arXiv 2024

  4. [4]

    DyCAST: Learning dynamic causal structure from time series

    Yue Cheng, Bochen Lyu, Weiwei Xing, and Zhanxing Zhu. DyCAST: Learning dynamic causal structure from time series. InInternational Conference on Learning Representations, 2025

    work page 2025

  5. [5]

    Eric J. Daza. Person as population: A longitudinal view of single-subject causal inference for analyzing self-tracked health data.arXiv preprint arXiv:1901.03423, 2019

    work page arXiv 1901

  6. [6]

    Foti, and Joseph Futoma

    Eray Erturk, Fahad Kamran, Salar Abbaspourazad, Sean Jewell, Harsh Sharma, Yujie Li, Sinead Williamson, Nicholas J. Foti, and Joseph Futoma. Beyond sensor data: Foundation models of behavioral data from wearables improve health predictions. InProceedings of the 42nd International Conference on Machine Learning, volume 267 ofProceedings of Machine Learning...

    work page 2025

  7. [7]

    C. W. J. Granger. Investigating causal relations by econometric models and cross-spectral methods.Econometrica, 37(3):424–438, 1969. doi: 10.2307/1912791

    work page doi:10.2307/1912791 1969

  8. [8]

    KCRL: A prior knowledge based causal discovery framework with reinforcement learning

    Uzma Hasan and Md Osman Gani. KCRL: A prior knowledge based causal discovery framework with reinforcement learning. InProceedings of the 7th Machine Learning for Healthcare Conference, pages 691–714, 2022

    work page 2022

  9. [9]

    Optimizing data-driven causal discovery using knowledge- guided search.arXiv preprint arXiv:2304.05493, 2024

    Uzma Hasan and Md Osman Gani. Optimizing data-driven causal discovery using knowledge- guided search.arXiv preprint arXiv:2304.05493, 2024

    work page arXiv 2024

  10. [10]

    DKC: Data-driven and knowledge-guided causal discovery with application to healthcare data.Knowledge-Based Systems, 337:115384, 2026

    Uzma Hasan and Md Osman Gani. DKC: Data-driven and knowledge-guided causal discovery with application to healthcare data.Knowledge-Based Systems, 337:115384, 2026. doi: 10. 1016/j.knosys.2026.115384. 10

    work page arXiv 2026

  11. [11]

    Bayesian nonparametric modeling for causal inference.Journal of Computa- tional and Graphical Statistics, 20(1):217–240, 2011

    Jennifer L Hill. Bayesian nonparametric modeling for causal inference.Journal of Computa- tional and Graphical Statistics, 20(1):217–240, 2011

    work page 2011

  12. [12]

    Johansson, Uri Shalit, and David Sontag

    Fredrik D. Johansson, Uri Shalit, and David Sontag. Learning representations for counterfactual inference. InProceedings of the 33rd International Conference on Machine Learning, volume 48 ofProceedings of Machine Learning Research, pages 3020–3029, 2016

    work page 2016

  13. [13]

    CDF-RAG: Causal dynamic feedback for adaptive retrieval-augmented generation.arXiv preprint arXiv:2504.12560, 2025

    Elahe Khatibi, Ziyu Wang, and Amir M Rahmani. CDF-RAG: Causal dynamic feedback for adaptive retrieval-augmented generation.arXiv preprint arXiv:2504.12560, 2025

    work page arXiv 2025

  14. [14]

    CARE-ECG: Causal Agent-based Reasoning for Explainable and Counterfactual ECG Interpretation

    Elahe Khatibi, Ziyu Wang, Ankita Sharma, Krishnendu Chakrabarty, Sanaz Rahimi Moosavi, Farshad Firouzi, and Amir Rahmani. CARE-ECG: Causal agent-based reasoning for explainable and counterfactual ecg interpretation.arXiv preprint arXiv:2604.10420, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  15. [15]

    CausalMed: Causality-based personalized medication recommendation centered on patient health state

    Xiang Li, Shunpan Liang, Yu Lei, Chen Li, Yulei Hou, Dashun Zheng, and Tengfei Ma. CausalMed: Causality-based personalized medication recommendation centered on patient health state. InProceedings of the 33rd ACM International Conference on Information and Knowledge Management, 2024

    work page 2024

  16. [16]

    BEHRT: Transformer for Electronic Health Records

    Yikuan Li, Shishir Rao, Jose Roberto Ayala Solares, Abdelaali Hassaine, Dexter Canoy, Yajie Zhu, Kazem Rahimi, and Gholamreza Salimi-Khorshidi. BEHRT: Transformer for electronic health records.Scientific Reports, 10(1):7155, 2020. doi: 10.1038/s41598-020-62922-y

    work page doi:10.1038/s41598-020-62922-y 2020

  17. [17]

    Forecasting treatment responses over time using recurrent marginal structural networks

    Bryan Lim, Ahmed Alaa, and Mihaela van der Schaar. Forecasting treatment responses over time using recurrent marginal structural networks. InAdvances in Neural Information Processing Systems, volume 31, pages 7493–7503, 2018

    work page 2018

  18. [18]

    Susan A. Murphy. Optimal dynamic treatment regimes.Journal of the Royal Statistical Society: Series B, 65(2):331–355, 2003

    work page 2003

  19. [19]

    Fairtabgen: Unifying counterfactual and causal fairness in synthetic tabular data generation.arXiv preprint arXiv:2508.11810, 2025

    Nitish Nagesh, Salar Shakibhamedan, Mahdi Bagheri, Ziyu Wang, Nima TaheriNejad, Axel Jantsch, and Amir M Rahmani. Fairtabgen: Unifying counterfactual and causal fairness in synthetic tabular data generation.arXiv preprint arXiv:2508.11810, 2025

    work page arXiv 2025

  20. [20]

    DYNOTEARS: Structure learning from time- series data

    Roxana Pamfil, Nisara Sriwattanaworachai, Shaan Desai, Philip Pilgerstorfer, Paul Beaumont, Konstantinos Georgatzis, and Bryon Aragam. DYNOTEARS: Structure learning from time- series data. InProceedings of the 23rd International Conference on Artificial Intelligence and Statistics, volume 108 ofProceedings of Machine Learning Research, pages 1595–1605, 2020

    work page 2020

  21. [21]

    Cambridge University Press, 2 edition, 2009

    Judea Pearl.Causality: Models, Reasoning, and Inference. Cambridge University Press, 2 edition, 2009

    work page 2009

  22. [22]

    Scalable and accurate deep learning with electronic health records,

    Alvin Rajkomar, Eyal Oren, Kai Chen, Andrew M. Dai, Nissan Hajaj, Michaela Hardt, Peter J. Liu, Xiaobing Liu, Jake Marcus, Mimi Sun, et al. Scalable and accurate deep learning with elec- tronic health records.npj Digital Medicine, 1(1):18, 2018. doi: 10.1038/s41746-018-0029-1

    work page doi:10.1038/s41746-018-0029-1 2018

  23. [23]

    et al.: Med-BERT: pretrained contextualized embed- dings on large-scale structured electronic health records for disease prediction

    Laila Rasmy, Yang Xiang, Ziqian Xie, Cui Tao, and Degui Zhi. Med-BERT: Pretrained contextualized embeddings on large-scale structured electronic health records for disease prediction.npj Digital Medicine, 4(1):86, 2021. doi: 10.1038/s41746-021-00455-y

    work page doi:10.1038/s41746-021-00455-y 2021

  24. [24]

    Robins, Miguel A

    James M. Robins, Miguel A. Hernan, and Babette Brumback. Marginal structural models and causal inference in epidemiology.Epidemiology, 11(5):550–560, 2000. doi: 10.1097/ 00001648-200009000-00011

    work page 2000

  25. [25]

    A causal discovery workflow for rare diseases: Experts-in-the-loop analysis of sparse longitudinal data.Journal of Medical Systems, 2026

    Niccolò Rocchi, Alessio Zanga, Alice Bernasconi, Annalisa Gronchi, Alessandroand Trama, Fabio Stella, et al. A causal discovery workflow for rare diseases: Experts-in-the-loop analysis of sparse longitudinal data.Journal of Medical Systems, 2026. doi: 10.1007/s10916-025-02327-4

    work page doi:10.1007/s10916-025-02327-4 2026

  26. [26]

    Detecting and quantifying causal associations in large nonlinear time series datasets

    Jakob Runge, Peer Nowack, Marlene Kretschmer, Seth Flaxman, and Dino Sejdinovic. Detecting and quantifying causal associations in large nonlinear time series datasets.Science Advances, 5 (11):eaau4996, 2019. doi: 10.1126/sciadv.aau4996. 11

    work page doi:10.1126/sciadv.aau4996 2019

  27. [27]

    Johansson, and David Sontag

    Uri Shalit, Fredrik D. Johansson, and David Sontag. Estimating individual treatment effect: Generalization bounds and algorithms. InProceedings of the 34th International Conference on Machine Learning, volume 70 ofProceedings of Machine Learning Research, pages 3076–3085, 2017

    work page 2017

  28. [28]

    Granger causality: A review and recent advances.Annual Review of Statistics and Its Application, 9:289–319, 2022

    Ali Shojaie and Emily B Fox. Granger causality: A review and recent advances.Annual Review of Statistics and Its Application, 9:289–319, 2022

    work page 2022

  29. [29]

    An algorithm for fast recovery of sparse causal graphs.Social Science Computer Review, 9(1):62–72, 1991

    Peter Spirtes and Clark Glymour. An algorithm for fast recovery of sparse causal graphs.Social Science Computer Review, 9(1):62–72, 1991. doi: 10.1177/089443939100900106

    work page doi:10.1177/089443939100900106 1991

  30. [30]

    Ajan Subramanian and Amir M. Rahmani. Personalized counterfactual framework: Generating potential outcomes from wearable data.arXiv preprint arXiv:2508.14432, 2025

    work page arXiv 2025

  31. [31]

    Alex Tank, Ian Covert, Nicholas Foti, Ali Shojaie, and Emily B. Fox. Neural granger causality. IEEE Transactions on Pattern Analysis and Machine Intelligence, 44(8):4267–4279, 2022. doi: 10.1109/TPAMI.2021.3065601

    work page doi:10.1109/tpami.2021.3065601 2022

  32. [32]

    Dynamic Expert-Guided Model Averaging for Causal Discovery

    Adrick Tench and Thomas Demeester. Dynamic expert-guided model averaging for causal discovery.arXiv preprint arXiv:2601.16715, 2026

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  33. [33]

    Time-varying vector autoregressive models with stochastic volatility

    Kostas Triantafyllopoulos. Time-varying vector autoregressive models with stochastic volatility. Journal of Applied Statistics, 38(2):369–382, 2011. doi: 10.1080/02664760903406512

    work page doi:10.1080/02664760903406512 2011

  34. [34]

    Differential private federated transfer learning for mental health monitoring in everyday settings: A case study on stress detection

    Ziyu Wang, Zhongqi Yang, Iman Azimi, and Amir M Rahmani. Differential private federated transfer learning for mental health monitoring in everyday settings: A case study on stress detection. In2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), pages 1–5. IEEE, 2024

    work page 2024

  35. [35]

    MedCoT-RAG: Causal chain-of-thought rag for medical question answering

    Ziyu Wang, Elahe Khatibi, and Amir M Rahmani. MedCoT-RAG: Causal chain-of-thought rag for medical question answering. In2025 IEEE 21st International Conference on Body Sensor Networks (BSN), pages 1–4. IEEE, 2025

    work page 2025

  36. [36]

    Healthq: Unveiling questioning capabilities of llm chains in healthcare conversations.Smart Health, 36: 100570, 2025

    Ziyu Wang, Hao Li, Di Huang, Hye-Sung Kim, Chae-Won Shin, and Amir M Rahmani. Healthq: Unveiling questioning capabilities of llm chains in healthcare conversations.Smart Health, 36: 100570, 2025

    work page 2025

  37. [37]

    Hang Wu, Wenqi Shi, and May D. Wang. Developing a novel causal inference algorithm for personalized biomedical causal graph learning using meta machine learning.BMC Medical Informatics and Decision Making, 24(1):137, 2024. doi: 10.1186/s12911-024-02510-6

    work page doi:10.1186/s12911-024-02510-6 2024

  38. [38]

    Coarse-to-fine learning of dynamic causal structures

    Dezhi Yang, Qiaoyu Tan, Carlotta Domeniconi, Jun Wang, Lizhen Cui, and Guoxian Yu. Coarse-to-fine learning of dynamic causal structures. InInternational Conference on Learning Representations, 2026

    work page 2026

  39. [39]

    GANITE: Estimation of individ- ualized treatment effects using generative adversarial nets

    Jinsung Yoon, James Jordon, and Mihaela van der Schaar. GANITE: Estimation of individ- ualized treatment effects using generative adversarial nets. InInternational Conference on Learning Representations, 2018

    work page 2018

  40. [40]

    DAG-GNN: DAG structure learning with graph neural networks

    Yue Yu, Jie Chen, Tian Gao, and Mo Yu. DAG-GNN: DAG structure learning with graph neural networks. InProceedings of the 36th International Conference on Machine Learning, volume 97 ofProceedings of Machine Learning Research, pages 7154–7163, 2019

    work page 2019

  41. [41]

    Ravikumar, and Eric P

    Xun Zheng, Bryon Aragam, Pradeep K. Ravikumar, and Eric P. Xing. DAGs with NO TEARS: Continuous optimization for structure learning. InAdvances in Neural Information Processing Systems, volume 31, pages 9472–9483, 2018. 12 A Additional Experimental Details A.1 Benchmark Construction The benchmark is designed to test the failure mode emphasized in the main...

    work page 2018