Pith. sign in

REVIEW 3 major objections 4 minor 36 references

Step-by-Step Guidance to Differential Anemia Diagnosis with Real-World Data and Deep Reinforcement Learning

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Deep reinforcement learning can guide anemia diagnosis test by test, matching batch classifiers while outputting an ordered pathway.

desk verdict A credible proof-of-concept that DRL can learn sequential test-request policies for anemia on real EHR data, but the headline 'competitive with SOTA' claim is undercut by post hoc selection of the best run/checkpoint in Table 2. read the letter →

arxiv 2412.02273 v1 pith:ZS2MRTGV submitted 2024-12-03 cs.LG

Lillian Muyama , Estelle Lu , Geoffrey Cheminet , Jacques Pouchot , Bastien Rance , Anne-Isabelle Tropeano , Antoine Neuraz , Adrien Coulet This is my paper · ORCID
classification cs.LG
keywords AnemiaDiagnosticpathwayClinicaldecisionsupportDeepQ-networkReinforcementlearningElectronichealthrecordsTransferDifferentialdiagnosis
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper tries to establish that a deep reinforcement learning agent can act like a clinician: instead of labeling a patient from all lab values at once, it requests one test at a time and stops when it can commit to a diagnosis. On a synthetic dataset generated from an expert-defined anemia decision tree, the agent nearly matches the oracle tree, with accuracy and F1 around 99 percent. On 1,127 real electronic health record patients, the best fine-tuned model reaches a macro-F1 (a class-averaged balance of precision and recall) of 63.48 and a ROC-AUC (a rank-based measure of class separation) of 77.67 while using an average of 3.19 tests per episode, outperforming random forests on F1 and coming close to the best batch classifier on ROC-AUC. The deeper claim is that a sequential strategy can be as accurate as batch classification while also being explainable, because every episode is an ordered list of requested tests. A secondary claim is that pre-training on synthetic data and fine-tuning on a small real cohort beats training on the real cohort alone.

What carries the argument

The load-bearing object is the Markov decision process formulation of diagnosis, together with the Dueling DQN-PER agent that solves it. In the MDP, the state is a fixed-size vector of feature values with -1 for unqueried features, the actions are either querying one of the 11 lab features or committing to a diagnosis, and the reward structure couples diagnostic accuracy with test count. Dueling DQN is a deep Q-network that splits the estimated value of a state into a state-value stream and an advantage stream for each action, and prioritized experience replay makes the agent sample the most informative past episodes; together they let the policy learn which test to request next from partial observations. The trajectory of actions is exactly the diagnostic pathway that the paper offers as the transparent counterpart to a batch classifier.

What would settle it

Take a random sample of about 200 of the 1,127 real-world patients, have two clinicians assign the final anemia sub-type from the full chart without seeing the model's output, and recompute the fine-tuned Dueling DQN-PER's F1 and ROC-AUC against that adjudicated gold standard; if the scores fall to chance level or clearly below the best batch classifier's, the paper's central claim is falsified.

Watch

Extended reading notes

Core claim

The authors formulate differential anemia diagnosis as a Markov decision process in which a patient's record is revealed feature by feature. At each step the state is the vector of lab values seen so far (with -1 marking values not yet requested); the possible actions are to ask for one of 11 anemia-related measurements or to declare one of the diagnosis classes; and the reward is +1 for a correct final diagnosis, -1 for a wrong one, and a small per-test penalty so that shorter pathways are preferred. Training a Dueling DQN with prioritized experience replay on 69,879 synthetic instances, then fine-tuning on the real cohort, produces a policy that, on the held-out real test set, achieves F1 63.48 and ROC-AUC 77.67 with mean episode length 3.19. The generated pathways are the interpretable core: the model cannot see all features at once, because it is rewarded for reaching the right class with the fewest tests. The authors interpret the gains on sickle cell anemia—recall rising from 9 percent on the synthetic-only model to 83 percent after fine-tuning, against 0 percent for the expert decision tree—as evidence that learned sequential policies can outperform a static guideline tree when the tree's thresholds or missing-value handling do not match real patients.

Load-bearing premise

The whole result depends on the electronic health record labels actually being confirmed anemia diagnoses; the authors' own evaluation of the matching algorithm used to create those labels found only 86 percent precision, with 42.9 percent of its errors being 'uncertain diagnoses' where the term was detected but did not mean a confirmed diagnosis.

Editorial extensions

If this is right

  • A clinician using this policy would receive not only a suggested diagnosis but a short, ordered list of tests, 3.19 on average in the best real-world model, so fewer unwarranted lab orders could be placed.
  • When real data are scarce, synthetic pre-training followed by fine-tuning is a workable route: the paper reports that this beats both training only on synthetic data and training only on the real cohort, and the improvement is largest for rare classes such as sickle cell anemia.
  • Sequential decision policies can beat static expert decision trees on real data: the expert-defined tree missed every sickle cell anemia case, while the fine-tuned model reached 83 percent recall, suggesting that learned thresholds handle missing and atypical values better.
  • Because the state and action space are defined only by a feature list and a diagnosis list, the same MDP should transfer to other lab-based differential diagnoses, which the authors name as their planned extension.
  • If these claims hold, an explainable sequential agent could complement batch classifiers in clinical decision support, preserving accuracy while adding step-by-step reasoning.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A direct extension the paper leaves implicit is to replace the uniform per-test penalty with real monetary or time costs for each lab test; the reward function already provides the hook, and this would turn 'shorter pathway' into 'cheaper or faster pathway.'
  • The paper's own limitation note that laboratory tests are usually ordered as panels suggests a panel-aware action space would change both the mean episode length and the clinical usefulness of the pathways the agent produces.
  • Because the manual note-adjudication set was used only to evaluate term detection, a natural follow-up is to adjudicate a sample of the final diagnosis labels themselves; this would reveal whether the reported F1 advantage over batch classifiers is real or an artifact of label noise.
  • The expert tree's failure on sickle cell anemia cases with missing ferritin suggests the learned policy is partly compensating for missing data; a complete-case analysis would separate missing-data handling from disease-specific learning.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper formulates anemia differential diagnosis as a sequential decision-making problem. It defines an MDP with feature-query actions and diagnostic actions, trains Dueling DQN and Dueling DDQN agents with Prioritized Experience Replay, and evaluates them on a synthetic dataset generated from an expert-defined decision tree and on a real-world dataset of 1,127 patients from a Paris hospital. Four training scenarios are compared: synthetic-only training, real-world-only training, direct transfer from synthetic to real data, and fine-tuning on real data. The authors report F1, ROC-AUC, accuracy, and mean episode length, and compare against random, tree-based, and standard machine-learning baselines. The central claim is that deep reinforcement learning performs competitively with state-of-the-art methods while additionally generating interpretable, step-by-step diagnostic pathways.

Significance. If the central claim were established, this would be a useful contribution to clinical decision support: it extends an earlier synthetic-only pipeline to a real-world clinical data warehouse, provides public code and data-description resources, and demonstrates interpretable sequential test selection with transfer learning. The transparent pathway generation is a genuine advantage over black-box classifiers, and the real-world evaluation with ICD-10 and NLP-derived labels is valuable despite label noise. However, the headline comparison in Table 2 is currently undermined by post hoc checkpoint selection and by zero-variance reporting for the selected DQN model, so the claim that DRL methods 'perform competitively' is not yet supported by the evidence as presented.

major comments (3)
  1. [§4.2 and Table D.3] The 'Best DQN' row in Table 2 is not a prespecified model: it is the maximum over three validation-selection criteria and ten seeds. The text states that the row is the best DQN in terms of F1 score, but Table D.3 shows the fine-tuned Dueling DQN-PER achieves test F1 = 63.48 when checkpoints are selected by ROC-AUC, while selection by F1 yields test F1 = 53.03 and selection by accuracy yields 51.75. The reported row is therefore the most favorable of several candidate selections, whereas the comparators (RF, XGBoost, FFNN) are standard tuned models reported as averages. This asymmetry invalidates the claimed comparison. The authors should prespecify a single checkpoint-selection metric and report the mean and standard deviation over the ten seeds under that fixed protocol, or alternatively report all three selection rows from D.3 as the primary comparison.
  2. [§5.1] The real-world ground-truth labels are derived from ICD-10 codes and a Medkit-based matcher with a reported precision of 86%, and 42.9% of the matcher's errors are 'uncertain diagnoses' where the detected term does not represent a confirmed diagnosis. Because these labels define both the reward signal used to train the DQN and the evaluation metrics in Table 2, the absolute F1 and ROC-AUC values are sensitive to this label noise. The authors acknowledge the limitation in Section 5.1 but do not quantify its impact. I request a robustness check on a manually verified random subset of the test set, or at least separate reporting of matcher precision for the test labels. Without such an analysis, the numerical performance of all models, including the DQN, is not fully reliable.
  3. [Tables D.2, D.3, and Table 2] The DQN rows in Tables D.2 and D.3 are reported with a standard deviation of ±0.00, while the same model class in Table D.1 shows F1 standard deviations of up to 6.69 points across ten seeds. This indicates that the DQN tables report one selected checkpoint per scenario rather than the distribution over the ten training runs. Zero-variance reporting obscures the run-to-run instability of the DQN method and undermines comparisons between training scenarios, such as the claimed improvement from fine-tuning. The authors should report mean and standard deviation over all ten runs for each selection metric, as they already do in Table D.1.
minor comments (4)
  1. [Author list] There is a typo in the author list: 'Anne-Isablle Tropeanoc' should be 'Anne-Isabelle Tropeano'.
  2. [§4.2 and Table D.2] The sentence 'models were selected based on their F1 scores, specifically those closest to the mean F1 score reported in Table 1' is ambiguous; it is unclear whether the selection is per seed, whether the chosen checkpoint is close to the mean validation F1 or the mean test F1, and how 'closest' is defined.
  3. [Table B.1 and §3.5] The paper does not report the total number of training timesteps or the checkpoint interval, although these are needed to understand how many checkpoints are evaluated; adding these values would improve reproducibility.
  4. [Table 2 caption] The baseline 'RWD-trained DT' in Table 2 is not defined in the table or its caption; the text later uses 'RWD-trained decision tree', so please make the abbreviation explicit.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity; the real-world evaluation is independent of the expert tree, and the DRL results do not reduce to fitted inputs.

full rationale

The paper's derivation chain is self-contained for its central claims. The synthetic dataset is generated from an expert-refined decision tree, and the tree-based agent's 100% score on that dataset is explicitly acknowledged as a consequence of construction: the authors state that the tree-based agent achieved a perfect score because it follows the expert-described DT used to label the dataset. This is therefore not presented as an independent prediction. The real-world evaluation uses labels derived from ICD-10 codes and Medkit-based note matching described in Section 3.3.2, which are independent of the DRL models and of the expert tree. The reward function in Equations 2 and 3 and the Q-learning updates in Appendix B are standard formulations and do not encode the target result. The only self-citation with any methodological weight is reference [3], which supplies the DRL architecture choices and the synthetic-data generation approach; this is prior-method transfer rather than a circular proof, and the headline real-world comparison does not reduce to that prior work. The limitations noted in Section 5.1, especially the 86% precision of the anemia-term matcher, are data-quality concerns that apply to all compared models and are not circularity. The Table 2 'Best DQN' row is selected across validation criteria and seeds, which is a legitimate concern about the strength of the state-of-the-art comparison, but it is not an equation-level reduction of a prediction to a fitted input. No circular step meeting the quoted-evidence bar was found.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central empirical claim rests on the MDP formulation, the reliability of automatically extracted diagnosis labels, and the validity of the synthetic data as a pretraining source. No fundamentally new physical or mathematical entities are introduced.

free parameters (4)
  • Feature query penalty coefficient = 1/(2n)
    Set by hand in Equation 3; controls the cost of each lab test and therefore the length of generated pathways.
  • Discount factor gamma = 0.99
    Chosen from prior knowledge and experimentation (Table B.1); affects how strongly future rewards are weighted.
  • Final exploration epsilon = 0.05
    Hyperparameter from Table B.1 that sets the exploration rate late in training.
  • Learning rate = 0.0001
    Optimizer learning rate from Table B.1, chosen by experimentation.
assumptions (5)
  • domain assumption The diagnostic process can be modeled as a Markov decision process where the state is the set of queried feature values and actions are test requests or terminal diagnoses.
    Section 3.1 defines the MDP; this assumes the sufficiency of the state representation for optimal decisions.
  • domain assumption ICD-10 codes and NLP-matched anemia terms in clinical notes are accurate enough to serve as ground truth for training and evaluation.
    Section 3.3.2 describes label construction; the paper reports 86% precision for the NLP matcher, indicating label noise.
  • domain assumption The expert-defined decision tree and its synthetic data capture the diagnostic logic for anemia subtypes.
    Section 3.3.1 uses this tree to generate 69,879 synthetic instances; the tree-following agent scores 100% on the synthetic test set by construction.
  • domain assumption The 11 laboratory and demographic features are sufficient to discriminate the anemia classes.
    Both datasets use the same feature set; the MDP only allows querying these features.
  • ad hoc to paper The reward function (correct diagnosis +1, incorrect -1, feature query -1/(2n)) induces clinically meaningful diagnostic behavior.
    Equations 2 and 3 in Section 3.1; this is a design choice not derived from clinical outcome data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Step-by-Step Guidance to Differential Anemia Diagnosis with Real-World Data and Deep Reinforcement Learning." pith.science (2026). https://pith.science/paper/ZS2MRTGV

@misc{pith2026241202273,
  author       = {Pith},
  title        = {Pith review of: Step-by-Step Guidance to Differential Anemia Diagnosis with Real-World Data and Deep Reinforcement Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ZS2MRTGV}},
  note         = {Machine review of arXiv:2412.02273}
}
read the original abstract

Clinical diagnostic guidelines outline the key questions to answer to reach a diagnosis. Inspired by guidelines, we aim to develop a model that learns from electronic health records to determine the optimal sequence of actions for accurate diagnosis. Focusing on anemia and its sub-types, we employ deep reinforcement learning (DRL) algorithms and evaluate their performance on both a synthetic dataset, which is based on expert-defined diagnostic pathways, and a real-world dataset. We investigate the performance of these algorithms across various scenarios. Our experimental results demonstrate that DRL algorithms perform competitively with state-of-the-art methods while offering the significant advantage of progressively generating pathways to the suggested diagnosis, providing a transparent decision-making process that can guide and explain diagnostic reasoning.

Figures

Figures reproduced from arXiv: 2412.02273 by the authors.

Figure 1
Figure 1. The expert-defined decision tree used to label the synthetic dataset. [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Diagnostic pathways generated by the fine-tuned dueling DQN-PER model. [PITH_FULL_IMAGE:figures/full_fig_p013_2.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 30 canonical work pages

  1. [3]

    Deep Reinforcement Learning for Personalized Diagnostic Decision Pathways Using Electronic Health Records: A Comparative Study on Anemia and Systemic Lupus Erythematosus

    L. Muyama, A. Neuraz, A. Coulet, Deep reinforcement learning for per- sonalizeddiagnosticdecisionpathwaysusingelectronichealthrecords: A comparative study on anemia and systemic lupus erythematosus, arXiv preprint arXiv:2404.05913 (2024)

    work page Pith review arXiv 2024

  2. [1]

    M. J. Field, K. N. Lohr, et al., Clinical practice guidelines, Directions for a new program (1990) 1990

    work page 1990

  3. [2]

    Wayfinding

    J. Adler-Milstein, J. H. Chen, G. Dhaliwal, Next-Generation Artifi- cial Intelligence for Diagnosis: From Predicting Diagnostic Labels to “Wayfinding”, JAMA 326 (24) (2021) 2467–2468.doi:10.1001/jama. 2021.22396. URL https://doi.org/10.1001/jama.2021.22396

    arXiv 2021

  4. [4]

    Bakhshi, E

    A. Bakhshi, E. Hassannayebi, A. H. Sadeghi, Optimizing sepsis care through heuristics methods in process mining: A trajectory analysis, Healthc. Anal. 3 (2023) 100187

    work page 2023

  5. [5]

    V. Vogt, S. M. Scholz, L. Sundmacher, Applying sequence clustering techniques to explore practice-based ambulatory care pathways in in- surance claims data, Eur. J. Public Health 28 (2) (2018) 214–219

    work page 2018

  6. [6]

    X. Xu, T. Jin, J. Wang, Summarizing patient daily activities for clinical pathway mining, in: 2016 IEEE 18th international conference on e- health networking, applications and services (Healthcom), IEEE, 2016, pp. 1–6

    work page 2016

  7. [7]

    Huang, Z

    Z. Huang, Z. Ge, W. Dong, K. He, H. Duan, Probabilistic modeling per- sonalizedtreatmentpathwaysusingelectronichealthrecords, J.Biomed. Inform. 86 (2018) 33–48

    work page 2018

  8. [8]

    Najjar, D

    A. Najjar, D. Reinharz, C. Girouard, C. Gagné, A two-step approach for mining patient treatment pathways in administrative healthcare databases, Artif. Intell. Med. 87 (2018) 34–48

    work page 2018

Show all 36 references
  1. [9]

    C. Hur, J. Wi, Y. Kim, Facilitating the development of deep learning models with visual analytics for electronic health records, Int. J. Envi- ron. Res. Public Health 17 (22) (2020) 8303. 16

    2020

  2. [10]

    Dagliati, V

    A. Dagliati, V. Tibollo, G. Cogni, L. Chiovato, R. Bellazzi, L. Sacchi, Careflow mining techniques to explore type 2 diabetes evolution, J. Di- abetes Sci. Technol. 12 (2) (2018) 251–259

    2018

  3. [11]

    Koshimizu, R

    H. Koshimizu, R. Kojima, K. Kario, Y. Okuno, Prediction of blood pressure variability using deep neural networks, International Journal of Medical Informatics 136 (2020) 104067

    2020

  4. [12]

    Miotto, L

    R. Miotto, L. Li, B. A. Kidd, J. T. Dudley, Deep patient: an unsuper- vised representation to predict the future of patients from the electronic health records, Scientific Reports 6 (1) (2016) 1–10

    2016

  5. [13]

    Z. C. Lipton, D. C. Kale, C. Elkan, R. Wetzel, Learning to diagnose with LSTM recurrent neural networks, arXiv preprint arXiv:1511.03677 (2015)

    2015 arXiv

  6. [14]

    E. Choi, M. T. Bahadori, A. Schuetz, W. F. Stewart, J. Sun, Doctor AI: Predicting clinical events via recurrent neural networks, in: Machine learning for healthcare conference, PMLR, 2016, pp. 301–318

    2016

  7. [15]

    6450–6459

    Y.Li, J.Oliva, Activefeatureacquisitionwithgenerativesurrogatemod- els, in: International Conference on Machine Learning, PMLR, 2021, pp. 6450–6459

    2021

  8. [16]

    Janisch, T

    J. Janisch, T. Pevn` y, V. Lis` y, Classification with costly features using deep reinforcement learning, in: Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 33, 2019, pp. 3959–3966

    2019

  9. [17]

    L. Huo, Y. Tang, Multi-objective deep reinforcement learning for per- sonalized dose optimization based on multi-indicator experience replay, Appl. Sci. 13 (1) (2022) 325

    2022

  10. [18]

    Maier, N

    C. Maier, N. Hartung, C. Kloft, W. Huisinga, J. de Wiljes, Reinforce- ment learning and Bayesian data assimilation for model-informed preci- sion dosing in oncology, CPT Pharmacometrics Syst. Pharmacol. 10 (3) (2021) 241–254

    2021

  11. [19]

    K.Gallagher, M.Strobl, R.Gatenby, P.Maini, A.Anderson, Learningto adapt-deep reinforcement learning in treatment-resistant prostate can- cer, bioRxiv (2023) 2023–04. 17

    2023

  12. [20]

    Moreau, V

    G. Moreau, V. François-Lavet, P. Desbordes, B. Macq, Reinforcement learning for radiotherapy dose fractioning automation, Biomedicines 9 (2) (2021) 214

    2021

  13. [21]

    T. Zhu, K. Li, P. Herrero, P. Georgiou, Basal glucose control in type 1 diabetes using deep reinforcement learning: An in silico validation, IEEE J. Biomed. Health Inform. 25 (4) (2020) 1223–1232

    2020

  14. [22]

    Z. Liu, L. Ji, X. Jiang, W. Zhao, X. Liao, T. Zhao, S. Liu, X. Sun, G. Hu, M. Feng, et al., A deep reinforcement learning approach for type 2 diabetes mellitus treatment, in: 2020 IEEE International Conference on Healthcare Informatics (ICHI), IEEE, 2020, pp. 1–9

    2020

  15. [23]

    Tang, H.-C

    K.-F. Tang, H.-C. Kao, C.-N. Chou, E. Y. Chang, Inquire and diagnose: Neural symptom checking ensemble using deep reinforcement learning, in: NIPS Workshop on Deep Reinforcement Learning, 2016

    2016

  16. [24]

    Z. Wei, Q. Liu, B. Peng, H. Tou, T. Chen, X.-J. Huang, K.-F. Wong, X. Dai, Task-oriented dialogue system for automatic diagnosis, in: Pro- ceedings of the 56th Annual Meeting of the Association for Computa- tional Linguistics (Volume 2: Short Papers), 2018, pp. 201–207

    2018

  17. [25]

    Kao, K.-F

    H.-C. Kao, K.-F. Tang, E. Chang, Context-aware symptom checking for disease diagnosis using hierarchical reinforcement learning, in: Proceed- ings of the AAAI Conference on Artificial Intelligence, Vol. 32, 2018

    2018

  18. [26]

    Z. Yu, Y. Li, J. Kim, K. Huang, Y. Luo, M. Wang, Deep rein- forcement learning for cost-effective medical diagnosis, arXiv preprint arXiv:2302.10261 (2023)

    2023 arXiv

  19. [27]

    Littman, Markov decision processes, in: N

    M. Littman, Markov decision processes, in: N. J. Smelser, P. B. Baltes (Eds.), International Encyclopedia of the Social & Behavioral Sciences, Pergamon, Oxford, 2001, pp. 9240–9242

    2001

  20. [28]

    R. S. Sutton, A. G. Barto, Reinforcement learning: An introduction, MIT press, 2018

    2018

  21. [29]

    C. J. Watkins, P. Dayan, Q-learning, Machine learning 8 (3) (1992) 279–292. 18

    1992

  22. [30]

    V. Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. Veness, M. G. Bellemare, A. Graves, M. Riedmiller, A. K. Fidjeland, G. Ostrovski, et al., Human-level control through deep reinforcement learning, Nature 518 (7540) (2015) 529–533

    2015

  23. [31]

    Neuraz, G

    A. Neuraz, G. Vaillant, C. Arias, O. Birot, K.-T. Huynh, T. Fabacher, A. Rogier, N. Garcelon, I. Lerner, B. Rance, A. Coulet, Facilitating phenotyping from clinical texts: the medkit library (2024). arXiv: 2409.00164. URL https://arxiv.org/abs/2409.00164

    2024 arXiv

  24. [32]

    Brockman, V

    G. Brockman, V. Cheung, L. Pettersson, J. Schneider, J. Schulman, J. Tang, W. Zaremba, OpenAI Gym, arXiv preprint arXiv:1606.01540 (2016)

    2016 arXiv

  25. [33]

    A. Hill, A. Raffin, M. Ernestus, A. Gleave, A. Kanervisto, R. Traore, P. Dhariwal, C. Hesse, O. Klimov, A. Nichol, M. Plappert, A. Radford, J. Schulman, S. Sidor, Y. Wu, Stable Baselines,https://github.com/ hill-a/stable-baselines (2018)

    2018

  26. [34]

    Van Hasselt, A

    H. Van Hasselt, A. Guez, D. Silver, Deep reinforcement learning with double Q-learning, in: Proceedings of the AAAI conference on artificial intelligence, Vol. 30, 2016

    2016

  27. [35]

    Z. Wang, T. Schaul, M. Hessel, H. Hasselt, M. Lanctot, N. Freitas, Du- eling network architectures for deep reinforcement learning, in: Interna- tional conference on machine learning, PMLR, 2016, pp. 1995–2003

    2016

  28. [36]

    Schaul, J

    T. Schaul, J. Quan, I. Antonoglou, D. Silver, Prioritized experience replay, arXiv preprint arXiv:1511.05952 (2015). 19 Appendix A. The environment We implemented our environment using the OpenAI Gym Python library. During training, the environment is reset using a random inst...

    2015 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.