pith. machine review for the scientific record. sign in

arxiv: 2604.17725 · v1 · submitted 2026-04-20 · 💻 cs.CL · cs.AI

Recognition: unknown

RePrompT: Recurrent Prompt Tuning for Integrating Structured EHR Encoders with Large Language Models

Arya Hadizadeh Moghaddam , Drew Ross , Mohsen Nayebi Kerdabadi , Dongjie Wang , Zijun Yao
Authors on Pith no claims yet

Pith reviewed 2026-05-10 04:46 UTC · model grok-4.3

classification 💻 cs.CL cs.AI
keywords prompt tuningelectronic health recordslarge language modelsclinical predictionrecurrent integrationMIMIClongitudinal data
0
0 comments X

The pith

RePrompT recurrently tunes LLM prompts with prior-visit latent states and cohort-derived population tokens to integrate structured EHR data without altering base models.

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

The paper proposes RePrompT as a way to feed time-stamped diagnosis and medication codes into large language models while keeping their original text-based strengths. It does this by feeding the LLM a running summary of past patient visits through recurrent latent vectors and by prepending a small set of learned prompt tokens that encode patterns seen across an entire patient population. The approach is tested on standard hospital datasets for tasks such as predicting death or readmission. A reader would care because most current LLM uses in medicine either lose the order of events when turning records into plain text or treat every patient in isolation, missing shared regularities that simpler EHR models already capture.

Core claim

RePrompT recurrently incorporates latent states from prior visits to preserve longitudinal information, and injects population-level information through trainable prompt tokens derived from a cohort-trained, task-aligned EHR encoder, without modifying the underlying LLM or EHR encoder architectures, and consistently outperforms both EHR-based and LLM-based baselines on MIMIC-III and MIMIC-IV across multiple clinical prediction tasks.

What carries the argument

Recurrent prompt tuning that injects time-aware latent states from previous visits together with population-level prompt tokens generated by a pre-trained EHR encoder.

If this is right

  • LLMs can retain temporal structure from sequential EHR codes without converting them to lossy text.
  • Population-level patterns learned by a separate encoder become available to the LLM during each patient's independent inference.
  • Existing LLMs and EHR encoders can be combined for clinical tasks without retraining either model from scratch.
  • Performance gains appear across mortality, readmission, and other standard prediction endpoints on large public ICU datasets.

Where Pith is reading between the lines

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

  • The same recurrent-plus-population prompt pattern could be tested on other time-ordered medical data such as vital-sign streams or imaging sequences.
  • If the prompt tokens prove stable across hospitals, they might serve as lightweight adapters that transfer between institutions without sharing raw patient records.
  • Removing the need to fine-tune the full LLM weights could lower the compute barrier for deploying such models in smaller clinical settings.

Load-bearing premise

Recurrent injection of latent states and population prompt tokens can be performed without modifying the LLM or EHR encoder while still capturing the necessary temporal order and code co-occurrence patterns.

What would settle it

RePrompT failing to outperform the strongest EHR-only and LLM-only baselines on at least two of the reported clinical prediction tasks when the recurrent state or population prompt components are ablated on MIMIC-III or MIMIC-IV.

Figures

Figures reproduced from arXiv: 2604.17725 by Arya Hadizadeh Moghaddam, Dongjie Wang, Drew Ross, Mohsen Nayebi Kerdabadi, Zijun Yao.

Figure 1
Figure 1. Figure 1: Illustration of the difference between exist [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The framework of the proposed method. Medical codes and discharge notes are first used to generate [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The Prompt for clinical record synthesis. [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Performance comparison of different EHR [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
read the original abstract

Large Language Models (LLMs) have shown strong promise for mining Electronic Health Records (EHRs) by reasoning over longitudinal clinical information to capture context-rich patient trajectories. However, leveraging LLMs for structured EHRs (e.g., standardized diagnosis and medication codes) presents two key challenges. First, translating time-stamped EHR sequences into plain text can obscure both temporal structure and code identities, weakening the ability to capture code co-occurrence and longitudinal regularities. Second, unlike cohort-trained predictive models that learn a shared, task-aligned representation space across patients, LLMs are often applied in a case-isolated inference setting where each patient is processed independently without leveraging population-level patterns. To address these challenges, we introduce RePrompT, a time-aware LLM framework that integrates structured EHR encoders through prompt tuning, without modifying underlying architectures. Specifically, RePrompT recurrently incorporates latent states from prior visits to preserve longitudinal information, and injects population-level information through trainable prompt tokens derived from a cohort-trained, task-aligned EHR encoder. Experiments on MIMIC-III and MIMIC-IV demonstrate that RePrompT consistently outperforms both EHR-based and LLM-based baselines across multiple clinical prediction tasks.

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

0 major / 3 minor

Summary. The paper introduces RePrompT, a framework for integrating structured EHR encoders with LLMs via recurrent prompt tuning without modifying either architecture. It recurrently feeds latent states from prior visits into the LLM as soft prompts to capture longitudinal information and derives additional trainable prompt tokens from a cohort-trained, task-aligned EHR encoder to inject population-level patterns. Experiments on MIMIC-III and MIMIC-IV are reported to show consistent outperformance over both EHR-only and LLM-only baselines on multiple clinical prediction tasks.

Significance. If the empirical results hold with proper controls and ablations, the work would be a useful engineering contribution to clinical NLP. It offers a lightweight integration strategy that preserves the strengths of both structured encoders (for co-occurrence and population statistics) and LLMs (for contextual reasoning) while addressing the loss of temporal structure when serializing EHR sequences to text. The approach could be adopted in settings where full fine-tuning of large models is impractical.

minor comments (3)
  1. The abstract asserts consistent outperformance on MIMIC-III/IV but supplies no quantitative metrics, baseline names, or error bars. Adding a compact results table or key numbers (e.g., AUROC deltas) would make the central claim immediately verifiable.
  2. The description of how latent states are mapped to prompt tokens and how recurrence is realized across visits (e.g., token concatenation order, masking, or state update rule) should be expanded with a diagram or pseudocode for reproducibility.
  3. Clarify whether the EHR encoder remains frozen during prompt tuning or is jointly optimized; the current wording leaves this ambiguous.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary of RePrompT and the recommendation for minor revision. The noted significance as a lightweight integration approach aligns with the paper's goals. No specific major comments were provided in the report.

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper presents an empirical engineering method (RePrompT) for integrating EHR encoders with LLMs via recurrent prompt tuning, with claims supported by experimental results on MIMIC-III/IV rather than any mathematical derivation. No load-bearing step reduces a prediction or result to its inputs by construction, self-definition, or self-citation chain; the core integration strategy (latent states turned into trainable prompts, recurrent injection into frozen LLM) is a described construction whose validity is tested externally via downstream task performance. The work is self-contained against benchmarks with no internal equivalence of output to fitted inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The framework rests on standard machine-learning assumptions about latent representations and prompt effectiveness rather than new axioms or invented entities.

free parameters (1)
  • trainable prompt tokens
    A small set of learned tokens derived from the cohort-trained EHR encoder and injected into LLM prompts; their values are fitted during the prompt-tuning stage.
axioms (2)
  • domain assumption Latent states from a structured EHR encoder can be recurrently passed forward to preserve longitudinal patient information without loss of critical temporal structure.
    Invoked when the paper states that recurrent incorporation of prior-visit states preserves longitudinal information.
  • domain assumption Population-level patterns learned by a cohort-trained EHR encoder can be transferred to an LLM via prompt tokens without requiring architectural changes to either model.
    Core premise of the integration method described in the abstract.

pith-pipeline@v0.9.0 · 5530 in / 1513 out tokens · 45625 ms · 2026-05-10T04:46:19.853660+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

123 extracted references · 20 canonical work pages · 8 internal anchors

  1. [1]

    Proceedings of the ACM Web Conference 2023 , pages=

    Molerec: Combinatorial drug recommendation with substructure-aware molecular representation learning , author=. Proceedings of the ACM Web Conference 2023 , pages=

    2023

  2. [2]

    proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Gamenet: Graph augmented memory networks for recommending medication combination , author=. proceedings of the AAAI Conference on Artificial Intelligence , volume=

  3. [3]

    arXiv preprint arXiv:2105.01876 , year=

    Change matters: Medication change prediction with recurrent residual networks , author=. arXiv preprint arXiv:2105.01876 , year=

    work page arXiv

  4. [4]

    Pediatric Research , volume=

    Using machine learning to impact on long-term clinical care: principles, challenges, and practicalities , author=. Pediatric Research , volume=. 2023 , publisher=

    2023

  5. [5]

    arXiv preprint arXiv:2105.02711 , year=

    Safedrug: Dual molecular graph encoders for recommending effective and safe drug combinations , author=. arXiv preprint arXiv:2105.02711 , year=

    work page arXiv

  6. [6]

    Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining , pages=

    Fair-srs: a fair session-based recommendation system , author=. Proceedings of the Fifteenth ACM International Conference on Web Search and Data Mining , pages=

  7. [7]

    Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining , pages=

    Efficiently leveraging multi-level user intent for session-based recommendation via atten-mixer network , author=. Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mining , pages=

  8. [8]

    Advances in neural information processing systems , volume=

    Retain: An interpretable predictive model for healthcare using reverse time attention mechanism , author=. Advances in neural information processing systems , volume=

  9. [9]

    The American journal of medicine , volume=

    Aspirin to prevent heart attack and stroke: what’s the right dose? , author=. The American journal of medicine , volume=. 2006 , publisher=

    2006

  10. [10]

    Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

    Adaptive disentangled transformer for sequential recommendation , author=. Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining , pages=

  11. [11]

    Scientific data , volume=

    MIMIC-III, a freely accessible critical care database , author=. Scientific data , volume=. 2016 , publisher=

    2016

  12. [12]

    JAMA Network Open , volume=

    Development and validation of a personalized model with transfer learning for acute kidney injury risk estimation using electronic health records , author=. JAMA Network Open , volume=. 2022 , publisher=

    2022

  13. [13]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Session-based recommendation with graph neural networks , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  14. [14]

    , author=

    Classification of the User's Intent Detection in Ecommerce systems-Survey and Recommendations. , author=. International Journal of Information Engineering & Electronic Business , volume=

  15. [15]

    Proceedings of the AAAI conference on artificial intelligence , volume=

    Learning the graphical structure of electronic health records with graph convolutional transformer , author=. Proceedings of the AAAI conference on artificial intelligence , volume=

  16. [17]

    Scientific reports , volume=

    Machine learning for comprehensive forecasting of Alzheimer’s Disease progression , author=. Scientific reports , volume=. 2019 , publisher=

    2019

  17. [18]

    Trends in pharmacological sciences , volume=

    The efficiency of multi-target drugs: the network approach might help drug design , author=. Trends in pharmacological sciences , volume=. 2005 , publisher=

    2005

  18. [19]

    Proceedings of the web conference 2021 , pages=

    Learning intents behind interactions with knowledge graph for recommendation , author=. Proceedings of the web conference 2021 , pages=

    2021

  19. [20]

    Proceedings of the ACM Web Conference 2022 , pages=

    Intent contrastive learning for sequential recommendation , author=. Proceedings of the ACM Web Conference 2022 , pages=

    2022

  20. [21]

    Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    based Multi-intention Contrastive Learning for Recommendation , author=. Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

  21. [22]

    International conference on machine learning , pages=

    A simple framework for contrastive learning of visual representations , author=. International conference on machine learning , pages=. 2020 , organization=

    2020

  22. [23]

    Agent: Clinical predictive model via mimicked second opinions , author=

    Dr. Agent: Clinical predictive model via mimicked second opinions , author=. Journal of the American Medical Informatics Association , volume=. 2020 , publisher=

    2020

  23. [24]

    , author=

    A convolutional net for medical records. , author=. 2017 , organization=

    2017

  24. [25]

    Australian Prescriber , volume=

    Antiemetic drugs: what to prescribe and when , author=. Australian Prescriber , volume=. 2020 , publisher=

    2020

  25. [26]

    Cough suppressant herbal drugs: A review , author=. Int. J. Pharm. Sci. Invent , volume=

  26. [27]

    The Pharmacologic Basis of Therapeutics

    Hypnotics and sedatives , author=. The Pharmacologic Basis of Therapeutics. 11th ed. Brunton LL, Lazo JS, Parker KL (Eds). New York, McGraw-Hill , pages=

  27. [28]

    American journal of hematology , volume=

    Intravenous iron: from anathema to standard of care , author=. American journal of hematology , volume=. 2008 , publisher=

    2008

  28. [29]

    ACM Transactions on Information Systems , volume=

    Knowledge-enhanced attributed multi-task learning for medicine recommendation , author=. ACM Transactions on Information Systems , volume=. 2023 , publisher=

    2023

  29. [30]

    Proceedings of the ACM Web Conference 2022 , pages=

    Conditional generation net for medication recommendation , author=. Proceedings of the ACM Web Conference 2022 , pages=

    2022

  30. [31]

    Advances in Neural Information Processing Systems , volume=

    Debiased, Longitudinal and Coordinated Drug Recommendation through Multi-Visit Clinic Records , author=. Advances in Neural Information Processing Systems , volume=

  31. [32]

    2018 IEEE international conference on data mining (ICDM) , pages=

    Self-attentive sequential recommendation , author=. 2018 IEEE international conference on data mining (ICDM) , pages=. 2018 , organization=

    2018

  32. [33]

    ACM Transactions on Information Systems , volume=

    Sequential recommendation with multiple contrast signals , author=. ACM Transactions on Information Systems , volume=. 2023 , publisher=

    2023

  33. [34]

    Proceedings of the 19th International Conference on World Wide Web , pages =

    Rendle, Steffen and Freudenthaler, Christoph and Schmidt-Thieme, Lars , title =. Proceedings of the 19th International Conference on World Wide Web , pages =. 2010 , isbn =

    2010

  34. [35]

    Proceedings of the 44th international ACM SIGIR conference on Research and development in information retrieval , pages=

    Augmenting sequential recommendation with pseudo-prior items via reversely pre-training transformer , author=. Proceedings of the 44th international ACM SIGIR conference on Research and development in information retrieval , pages=

  35. [36]

    2020 IEEE International Conference on Big Data (Big Data) , pages=

    Basket recommendation with multi-intent translation graph neural network , author=. 2020 IEEE International Conference on Big Data (Big Data) , pages=. 2020 , organization=

    2020

  36. [37]

    IEEE Transactions on Multimedia , volume=

    Hierarchical user intent graph network for multimedia recommendation , author=. IEEE Transactions on Multimedia , volume=. 2021 , publisher=

    2021

  37. [38]

    Proceedings of The Web Conference 2020 , pages=

    Attentive sequential models of latent intent for next item recommendation , author=. Proceedings of The Web Conference 2020 , pages=

    2020

  38. [39]

    Proceedings of the 32th international joint conference on artificial intelligence , pages=

    Intent-aware recommendation via disentangled graph contrastive learning , author=. Proceedings of the 32th international joint conference on artificial intelligence , pages=

  39. [40]

    Advances in neural information processing systems , volume=

    Pytorch: An imperative style, high-performance deep learning library , author=. Advances in neural information processing systems , volume=

  40. [41]

    Proceedings of the 27th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD) 2023 , year =

    Yang, Chaoqi and Wu, Zhenbang and Jiang, Patrick and Lin, Zhen and Gao, Junyi and Danek, Benjamin and Sun, Jimeng , title =. Proceedings of the 27th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD) 2023 , year =

    2023

  41. [42]

    Advances in neural information processing systems , volume=

    Are sixteen heads really better than one? , author=. Advances in neural information processing systems , volume=

  42. [43]

    The American journal of pathology , volume=

    Dietary iron overload induces visceral adipose tissue insulin resistance , author=. The American journal of pathology , volume=. 2013 , publisher=

    2013

  43. [44]

    Current nutrition reports , volume=

    The role of vitamin K in chronic aging diseases: inflammation, cardiovascular disease, and osteoarthritis , author=. Current nutrition reports , volume=. 2016 , publisher=

    2016

  44. [45]

    Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    Multi-level cross-view contrastive learning for knowledge-aware recommender system , author=. Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

  45. [46]

    Advances in Neural Information Processing Systems , volume=

    REFINE: A Fine-Grained Medication Recommendation System Using Deep Learning and Personalized Drug Interaction Modeling , author=. Advances in Neural Information Processing Systems , volume=

  46. [47]

    Nature Communications , volume=

    TransformEHR: transformer-based encoder-decoder generative model to enhance prediction of disease outcomes using electronic health records , author=. Nature Communications , volume=. 2023 , publisher=

    2023

  47. [48]

    arXiv preprint arXiv:2305.12788 , year=

    Graphcare: Enhancing healthcare predictions with personalized knowledge graphs , author=. arXiv preprint arXiv:2305.12788 , year=

    work page arXiv

  48. [49]

    Proceedings of the 33rd ACM International Conference on Information and Knowledge Management , pages=

    Contrastive learning on medical intents for sequential prescription recommendation , author=. Proceedings of the 33rd ACM International Conference on Information and Knowledge Management , pages=

  49. [50]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    ConSequence: Synthesizing Logically Constrained Sequences for Electronic Health Record Generation , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  50. [51]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    VITA:‘Carefully Chosen and Weighted Less’ Is Better in Medication Recommendation , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  51. [52]

    Advances in Neural Information Processing Systems , volume=

    REFINE: a fine-grained medication recommendation system using deep learning and personalized drug interaction modeling , author=. Advances in Neural Information Processing Systems , volume=

  52. [53]

    Proceedings of the ACM Web Conference 2023 , pages=

    Seqcare: Sequential training with external medical knowledge graph for diagnosis prediction in healthcare data , author=. Proceedings of the ACM Web Conference 2023 , pages=

    2023

  53. [54]

    Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining , pages=

    GRAM: graph-based attention model for healthcare representation learning , author=. Proceedings of the 23rd ACM SIGKDD international conference on knowledge discovery and data mining , pages=

  54. [55]

    Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    Metacare++: Meta-learning with hierarchical subtyping for cold-start diagnosis prediction in healthcare data , author=. Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

  55. [56]

    Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages=

    Hierarchical attention propagation for healthcare representation learning , author=. Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages=

  56. [57]

    Proceedings of the 27th ACM international conference on information and knowledge management , pages=

    Kame: Knowledge-based attention model for diagnosis prediction in healthcare , author=. Proceedings of the 27th ACM international conference on information and knowledge management , pages=

  57. [58]

    Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    Warm up cold-start advertisements: Improving ctr predictions via learning to learn id embeddings , author=. Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

  58. [59]

    Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

    Meta-learning on heterogeneous information networks for cold-start recommendation , author=. Proceedings of the 26th ACM SIGKDD international conference on knowledge discovery & data mining , pages=

  59. [60]

    Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages=

    Melu: Meta-learned user preference estimator for cold-start recommendation , author=. Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining , pages=

  60. [61]

    Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    M2eu: Meta learning for cold-start recommendation via enhancing user preference estimation , author=. Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

  61. [62]

    , author=

    Preference-adaptive meta-learning for cold-start recommendation. , author=. IJCAI , pages=

  62. [63]

    PhysioNet

    Mimic-iv , author=. PhysioNet. Available online at: https://physionet. org/content/mimiciv/1.0/(accessed August 23, 2021) , pages=

    2021

  63. [64]

    Advances in Neural Information Processing Systems , year=

    Attention is all you need , author=. Advances in Neural Information Processing Systems , year=

  64. [65]

    Companion Proceedings of the ACM on Web Conference 2024 , pages=

    Large Language Models as Data Augmenters for Cold-Start Item Recommendation , author=. Companion Proceedings of the ACM on Web Conference 2024 , pages=

    2024

  65. [66]

    Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

    Privileged graph distillation for cold start recommendation , author=. Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages=

  66. [67]

    Proceedings of the 29th ACM International Conference on Multimedia , pages=

    Contrastive learning for cold-start recommendation , author=. Proceedings of the 29th ACM International Conference on Multimedia , pages=

  67. [68]

    International conference on machine learning , pages=

    Model-agnostic meta-learning for fast adaptation of deep networks , author=. International conference on machine learning , pages=. 2017 , organization=

    2017

  68. [69]

    Scientific data , volume=

    MIMIC-IV, a freely accessible electronic health record dataset , author=. Scientific data , volume=. 2023 , publisher=

    2023

  69. [70]

    IEEE journal of biomedical and health informatics , volume=

    Deepr: a convolutional net for medical records , author=. IEEE journal of biomedical and health informatics , volume=. 2016 , publisher=

    2016

  70. [71]

    BMC Medical Informatics and Decision Making , volume=

    KAMPNet: multi-source medical knowledge augmented medication prediction network with multi-level graph contrastive learning , author=. BMC Medical Informatics and Decision Making , volume=. 2023 , publisher=

    2023

  71. [72]

    2019 International Joint Conference on Neural Networks (IJCNN) , pages=

    Meta-learning for user cold-start recommendation , author=. 2019 International Joint Conference on Neural Networks (IJCNN) , pages=. 2019 , organization=

    2019

  72. [73]

    Journal of biomedical informatics , volume=

    Deep learning for temporal data representation in electronic health records: A systematic review of challenges and methodologies , author=. Journal of biomedical informatics , volume=. 2022 , publisher=

    2022

  73. [74]

    Journal of Medical Internet Research , volume=

    The use of deep learning and machine learning on longitudinal electronic health records for the early detection and prevention of diseases: scoping review , author=. Journal of Medical Internet Research , volume=. 2024 , publisher=

    2024

  74. [75]

    , author=

    Medical Concept Embedding with Multiple Ontological Representations. , author=. IJCAI , volume=

  75. [76]

    Lihong Song, Chin Wang Cheong, Kejing Yin, William K Cheung, Benjamin CM Fung, and Jonathan Poon

    Pre-training of graph augmented transformers for medication recommendation , author=. arXiv preprint arXiv:1906.00346 , year=

    work page arXiv 1906

  76. [77]

    ACM Transactions on Knowledge Discovery from Data , year=

    Discovering Time-Aware Hidden Dependencies with Personalized Graphical Structure in Electronic Health Records , author=. ACM Transactions on Knowledge Discovery from Data , year=

  77. [78]

    Proceedings of the 32nd ACM International Conference on Information and Knowledge Management , pages=

    Contrastive learning of temporal distinctiveness for survival analysis in electronic health records , author=. Proceedings of the 32nd ACM International Conference on Information and Knowledge Management , pages=

  78. [79]

    international conference on machine learning , pages=

    Dropout as a bayesian approximation: Representing model uncertainty in deep learning , author=. international conference on machine learning , pages=. 2016 , organization=

    2016

  79. [80]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume=

    Building calibrated deep models via uncertainty matching with auxiliary interval predictors , author=. Proceedings of the AAAI Conference on Artificial Intelligence , volume=

  80. [81]

    Advances in neural information processing systems , volume=

    Simple and scalable predictive uncertainty estimation using deep ensembles , author=. Advances in neural information processing systems , volume=

Showing first 80 references.