Pith. sign in

REVIEW 2 major objections 4 minor 64 references

Deep Reinforcement Learning based Triggering Function for Early Classifiers of Time Series

T0 review · 2 major / 4 minor · reviewed 2026-08-08 · deepseek-v4-flash

Pith's one-line read Alert*, a reinforcement-learned trigger, achieves lower average misclassification-plus-delay cost than state-of-the-art early classifiers across all cost weightings on 31 datasets.

desk verdict A mostly solid empirical paper with a genuine result for high delay costs, undercut by an overstated conclusion and a flawed 'same information' comparison for one baseline. read the letter →

arxiv 2502.06584 v1 pith:PHKQ7V7P submitted 2025-02-10 cs.LG

classification cs.LG
keywords earlyclassificationoftimeseriesreinforcementlearningtriggeringfunctiondeepQ-networkmisclassificationcostdelayoffline
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

Early classification of time series requires a trigger that decides, before the whole series is seen, whether to wait or to predict. The paper asks whether the hand-coded rules used by existing triggers are actually good, and answers by recasting the trigger as a reinforcement learning problem with the same input features. Fair comparisons on 31 datasets show that hand-coded rules remain competitive when the state space is small, but that an RL trigger fed a combination of classifier features, Alert*, achieves lower average misclassification-plus-delay cost than all tested competitors for every cost weighting, significantly so when delay costs dominate.

What carries the argument

The central machinery is the decomposition of a triggering function into a criterion c and a decision rule r, together with the replacement of r by a learned policy. The RL formulation defines states as the classifier's feature set F, actions as wait or trigger, and rewards as the negative incremental delay cost, plus the negative misclassification cost when the trigger fires. A Double Deep Q-Network (DDQN) approximates the optimal Q-function, and its argmax policy decides when to stop. Alert* is the variant whose state space combines five features: predicted class, maximum posterior, margin, confidence-level bin, and elapsed time.

What would settle it

Retrain the Economy RL counterpart with a state that includes the full vector of expected future costs computed from the posterior probabilities and the cost functions, on the same 31 datasets and cost settings, and compare against Economy. If the enriched RL agent matches or beats Economy, the claim that man-tailored rules dominate on small state spaces fails; if it still loses, the claim survives.

Watch

Extended reading notes

Core claim

The paper's central claim is that a triggering function can be learned instead of designed, and that the learned version can beat the man-tailored rules. Using a Double Deep Q-Network trained offline with rewards equal to the negative incremental delay cost and, on triggering, the misclassification cost, the authors build Alert*, whose state space combines the predicted class, the maximum posterior probability, the margin, a binned confidence level, and the elapsed time. In experiments over 31 public datasets, Alert* dominates all state-of-the-art methods on average for all values of the trade-off parameter alpha, with statistically significant wins for alpha greater than 0.5, while the hand-tailored Economy rule beats its RL counterpart when both use a very small state space.

Load-bearing premise

The whole comparison rests on each RL agent seeing exactly the same information as its man-tailored counterpart, yet for Economy the RL state contains only a binned confidence level while the Economy rule computes predicted cost expectations from posterior probabilities and the cost functions; if that information gap is real, the conclusion that hand-coded rules win on small state spaces is not established.

Editorial extensions

If this is right

  • Any separable early-classification system can upgrade its trigger by training a DDQN on the features its classifier already outputs, without changing the classifier.
  • The fair-comparison protocol gives the field a way to test whether a proposed hand-designed trigger actually improves on a learned policy given the same information.
  • Larger, well-chosen state spaces improve learned triggers; adding raw time-series values or random noise does not, so feature choice matters more than raw data.
  • At low delay costs every method triggers immediately and differences vanish, so gains from learned triggers concentrate in regimes where waiting has a real price.
  • End-to-end RL methods like Earliest remain competitive with strong baselines despite seeing cost only through a hyperparameter, suggesting they deserve further study.

Reading between the lines

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

  • The Economy comparison may be unfair: its RL counterpart's state is only a binned confidence level, while the Economy rule uses predicted cost expectations computed from posterior probabilities and the cost functions; giving the RL agent those expectations could erase or reverse the apparent advantage of the hand-coded rule.
  • The success of Alert* suggests a general recipe: for any separable ECTS method, define the state as the classifier's full output distribution plus time, train offline, and obtain a trigger that is at least as good as the hand-designed one; this could be tested on multivariate series or with non-additive costs.
  • The paper's finding that a conservative trigger (waiting past the a posteriori best time) can be profitable is a testable hypothesis about robustness to label noise: on cleaner datasets the optimal policy should approach the a posteriori best trigger time more closely.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 4 minor

Summary. The paper addresses two questions in early classification of time series (ECTS): (1) whether reinforcement-learning-based triggering rules can outperform man-tailored decision rules when both use the same classifier and the same feature set, and (2) whether a larger, hand-selected feature set improves an RL-based trigger. The authors formulate separable ECTS as a reinforcement learning problem with a reward equal to the negative of the misclassification-plus-delay cost, instantiate RL counterparts for Stopping Rule, Economy, Calimera, and Proba Threshold, and compare them on 31 datasets. They then introduce Alert*, an RL trigger whose state space combines all features used by these baselines, and report that it has the best mean rank over the whole alpha range, with statistical significance for alpha > 0.5. The paper concludes that Alert* significantly outperforms state-of-the-art competitors for all weighted combinations of the two costs.

Significance. If the main empirical result holds, the paper makes a useful contribution: it provides a reusable RL formulation of separable ECTS triggering, a controlled comparison protocol with a shared classifier, and a strong new trigger, Alert*, supported by Wilcoxon tests and bootstrap confidence intervals. The code is released, which is a concrete strength. The controlled comparison for the Economy baseline, however, is not actually controlled, and the conclusion overstates the significance of the Alert* result. These issues are fixable and do not invalidate the main Alert* versus state-of-the-art comparison, but they must be addressed before the paper can be accepted.

major comments (2)
  1. [Section 7 and Appendix A.3, Table 2] The conclusion states that Alert* 'significantly outperformed its state-of-the-art competitors for all weighted combinations' of misclassification and delay costs, but this is contradicted by the paper's own Table 2: for alpha <= 0.5, many p-values are above 0.05, and Section 6.5.2 explicitly acknowledges that differences for alpha <= 0.5 are not significant. The conclusion should be restricted to the range alpha > 0.5 or rephrased to report the actual significance pattern over alpha.
  2. [Section 6.3 and Figure 3a] The claim that each RL counterpart receives 'exactly the same information' as its man-tailored baseline is not satisfied for Economy. Economy's rule compares expected costs computed from posterior probabilities and both cost functions over current and future times (Section 4.1), whereas Alert_Eco's state is only {level of confidence} (Figure 3a), a binned maximum posterior that does not include the current time t. Since the reward in Eq. (7) contains delay-cost increments that depend on t, the MDP is not Markovian without time in the state, and the RL agent cannot recover the delay increment it pays when choosing 'wait'. The resulting comparison does not isolate the decision rule, and the conclusion that man-tailored rules beat their RL counterparts when the input space is small is not established by this experiment. Re-running Alert_Eco with Economy's actual inputs, including t and the expected-cost features, would settle the question.
minor comments (4)
  1. [Section 6.1] The dataset count is inconsistent: the text says '31 datasets: 20 from the UCR archive and 114 from the Monash time series extrinsic regression archive', which cannot sum to 31; this should be corrected to 11 (or the selection procedure should be clarified).
  2. [Figure 4b caption] The caption says the Wilcoxon signed-rank test is 'labeled with mean A vgCost', which is unclear; it should state whether the labels are mean ranks or mean costs, and how they relate to the displayed test.
  3. [Section 5.4] The model selection procedure is described only verbally; a pseudocode or a precise statement of the epoch grid, the number of splits, and the criterion for 'best-performing one over all splits' would improve reproducibility.
  4. [General] The notation Alert* (with asterisk) is used in Section 5.3 and later, but Figure 4a and the conclusion sometimes render it as 'Alert⋆'; please unify the notation.

Circularity Check

0 steps flagged · score 1.0 of 10

No material circularity: the RL reward is the same cost the evaluation measures, which is a legitimate optimization objective; the Alert* result rests on an independent empirical comparison.

full rationale

None of the claimed results reduces to its inputs by construction. The reward in Eq. (7) is the negative of the same misclassification and delay costs appearing in the evaluation metric Eq. (11); that is a deliberate training objective, not a circular prediction, and all compared methods are scored with the same AvgCost metric. The headline Alert* versus SOTA comparison is an empirical benchmark computed in this paper against external algorithms (Calimera, Earliest, Proba Threshold, Stopping Rule, Economy), with Wilcoxon tests reported in Appendix A.3; it does not import its conclusion from the authors' earlier work. Self-citations ([36], [1], [6], [45], [53]) are used to select competitors and define the Economy baseline, not to assert Alert*'s superiority. The one noteworthy validity threat is the Sec. 6.3 claim that RL counterparts receive "exactly the same information": for Economy, the man-tailored rule triggers on expected costs computed from posteriors and cost functions (Sec. 4.1), while Alert_Eco's state is only {level of confidence} (Fig. 3a), so that pairwise comparison may understate what RL could do with Economy's inputs. That is a fairness/correctness issue, not a circular reduction, and it does not affect the independence of the main empirical result. Overall, no significant circularity is present.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim depends on the RL training setup and the chosen state space. The main free parameters are RL hyperparameters that are not fully reported, and the axioms concern the additive cost structure, the sufficiency of the feature set, and the fixed classifier. No new physical or conceptual entities are introduced.

free parameters (5)
  • discount factor gamma = not reported
    Equation (6) includes a discount factor gamma, but its value is not stated. If gamma < 1, the RL objective does not exactly match the total misclassification and delay cost that the evaluation metric measures.
  • number of bins for confidence discretization = not reported
    The 'level of confidence' feature is defined as a bin index within an equal-frequency discretization of maximum posteriors (Section 5.3). The number of bins is not specified, and it directly shapes the state representation.
  • number of train/validation splits = not reported
    Section 5.4 says several train/validation splits are used for model selection, but the exact number is not given, making the selection procedure incompletely specified.
  • Q-network hidden dimension = 32
    Section 6.4 sets the hidden dimension of the single-layer Q-network to 32. This is a hand-chosen architectural parameter.
  • Adam learning rate and target update tau = 1e-4 and 3e-3
    Section 6.4 specifies these DDQN hyperparameters, which are chosen by hand and affect the training dynamics and final policy.
assumptions (3)
  • domain assumption The misclassification and delay costs are additively decomposable and known to the agent.
    The reward function in Equation (7) is built from C_m and C_d separately, and Section 5.2 explicitly notes the additive decomposition assumption, with an extension to delayed rewards in Appendix A.5.
  • domain assumption The chosen state features are sufficient for near-optimal triggering decisions.
    Alert* uses a fixed set of features (predicted class, max posterior, margin, confidence bin, time) as the complete state. If an important piece of information, such as raw series shape, were necessary for optimal triggering, the learned policy would be suboptimal. Section 5.3 defines this feature set.
  • domain assumption The classifier h is fixed, well-calibrated, and its outputs capture all relevant classification information.
    The methodology separates the classifier from the trigger, and all comparisons use MiniROCKET with a calibration step (Section 6.4). If the classifier were poorly calibrated, the state features would be less informative.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep Reinforcement Learning based Triggering Function for Early Classifiers of Time Series." pith.science (2026). https://pith.science/paper/PHKQ7V7P

@misc{pith2026250206584,
  author       = {Pith},
  title        = {Pith review of: Deep Reinforcement Learning based Triggering Function for Early Classifiers of Time Series},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PHKQ7V7P}},
  note         = {Machine review of arXiv:2502.06584}
}
read the original abstract

Early Classification of Time Series (ECTS) has been recognized as an important problem in many areas where decisions have to be taken as soon as possible, before the full data availability, while time pressure increases. Numerous ECTS approaches have been proposed, based on different triggering functions, each taking into account various pieces of information related to the incoming time series and/or the output of a classifier. Although their performances have been empirically compared in the literature, no studies have been carried out on the optimality of these triggering functions that involve ``man-tailored'' decision rules. Based on the same information, could there be better triggering functions? This paper presents one way to investigate this question by showing first how to translate ECTS problems into Reinforcement Learning (RL) ones, where the very same information is used in the state space. A thorough comparison of the performance obtained by ``handmade'' approaches and their ``RL-based'' counterparts has been carried out. A second question investigated in this paper is whether a different combination of information, defining the state space in RL systems, can achieve even better performance. Experiments show that the system we describe, called \textsc{Alert}, significantly outperforms its state-of-the-art competitors on a large number of datasets.

Figures

Figures reproduced from arXiv: 2502.06584 by the authors.

Figure 1
Figure 1. Different architectures for the ECTS problem. The [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Heatmap representing decision rule 𝑟 on the Chilled￾WaterPredictor dataset, learned by Stopping Rule (2a) and using RL (2b), based on (i) the maximum probability esti￾mated by ℎ, in 𝑦-axis and (ii) the proportion seen of the time series, in 𝑥-axis (see Section 6). Red lines delimit areas where the probability of triggering, estimated by a sigmoid func￾tion, is above 0.5. (1) First, we present a methodology to transl… view at source ↗
Figure 3
Figure 3. Pairwise comparison of SOTA methods versus their [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The ranking plot (a) shows that, across all [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Pareto front, displaying for each 1 𝛼, the normal￾ized version of the AvgCost, decomposed over delay and mis￾classification cost on 𝑥-axis and 𝑦-axis respectively. Best ap￾proaches are located on the top left corner. High 𝛼 values are located on the right, low ones on …
Figure 6
Figure 6. Figure 6: The 𝑥-axis reports how far is the triggering time from the best a posteriori one: left is better. The 𝑦-axis reports the difference between the AvgCost incurred by the algorithm compared to the best a posteriori one, AvgCost★ : lower is better. The black crosses report…
Figure 7
Figure 7. Figure 7: shows that the original version of Alert★ is still the best performing algorithm, even if not significantly better than Alert★&series, except for 𝛼 = 0.9 (see Appendix A.6). This demon￾strates that the state space of Alert★ is well chosen and that adding more unprocess…
Figure 8
Figure 8. Figure 8: Pairwise comparison of Proba Threshold vs. RL counterpart using same information as input. Points above the horizontal line indicates that the man-tailored method is better than its RL-based counterpart A.2 Pairwise comparison: statistical tests [PITH_FULL_IMAGE:figur…
Figure 10
Figure 10. Figure 10: Evolution of the mean ranks, for every 𝛼, based on the AvgCost metric. A.6 State space study: statistical tests [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

64 extracted references · 55 canonical work pages

  1. [1]

    Youssef Achenchabe, Alexis Bondu, Antoine Cornuéjols, and Asma Dachraoui

  2. [2]

    Anthony Bagnall, Jason Lines, Aaron Bostrom, James Large, and Eamonn Keogh

  3. [3]

    Philip J Ball, Laura Smith, Ilya Kostrikov, and Sergey Levine. 2023. Efficient online reinforcement learning with offline data. In International Conference on Machine Learning. PMLR, 1577–1594

  4. [4]

    Jakub Michal Bilski and Agnieszka Jastrzebska. 2023. CALIMERA: A new early time series classification method. Information Processing & Management 60, 5 (2023), 103465

  5. [5]

    Alexis Bondu, Youssef Achenchabe, Albert Bifet, Fabrice Clérot, Antoine Cor- nuéjols, Joao Gama, Georges Hébrail, Vincent Lemaire, and Pierre-François Marteau. 2022. Open challenges for machine learning based early decision- making research. ACM SIGKDD Explorations Newsletter 24, 2 (2022), 12–31

  6. [6]

    Asma Dachraoui, Alexis Bondu, and Antoine Cornuéjols. 2015. Early classification of time series as a non myopic sequential decision mak- ing problem. In Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2015, Porto, Portugal, September 7-11, 2015, Proceedings, Part I 15. Springer, 433–447

  7. [7]

    Hoang Anh Dau, Anthony Bagnall, Kaveh Kamgar, Chin-Chia Michael Yeh, Yan Zhu, Shaghayegh Gharghabi, Chotirat Ann Ratanamahatana, and Eamonn Keogh

  8. [8]

    Peter Dayan and CJCH Watkins. 1992. Q-learning. Machine learning 8, 3 (1992), 279–292

Show all 64 references
  1. [9]

    Angus Dempster, Daniel F Schmidt, and Geoffrey I Webb. 2021. Minirocket: A very fast (almost) deterministic transform for time series classification. In Proceedings of the 27th ACM SIGKDD conference on knowledge discovery & data mining. 248–257

  2. [10]

    Akinori F Ebihara, Taiki Miyagawa, Kazuyuki Sakurai, and Hitoshi Imaoka. 2020. Sequential density ratio estimation for simultaneous optimization of speed and accuracy. arXiv preprint arXiv:2006.05587 (2020)

  3. [11]

    Akinori F Ebihara, Taiki Miyagawa, Kazuyuki Sakurai, and Hitoshi Imaoka

  4. [12]

    Scott Fujimoto and Shixiang Shane Gu. 2021. A minimalist approach to offline reinforcement learning. Advances in neural information processing systems 34 (2021), 20132–20145

  5. [13]

    Thomas Hartvigsen, Walter Gerych, Jidapa Thadajarassiri, Xiangnan Kong, and Elke Rundensteiner. 2022. Stop&hop: Early classification of irregular time series. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management. 696–705

  6. [14]

    Thomas Hartvigsen, Cansu Sen, Xiangnan Kong, and Elke Rundensteiner. 2019. Adaptive-halting policy network for early classification. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 101–110

  7. [15]

    Thomas Hartvigsen, Cansu Sen, Xiangnan Kong, and Elke Rundensteiner. 2020. Recurrent halting chain for early multi-label classification. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. 1382–1392

  8. [16]

    Yu Huang, Gary G Yen, and Vincent S Tseng. 2022. Snippet policy network for multi-class varied-length ECG early classification. IEEE Transactions on Knowledge and Data Engineering 35, 6 (2022), 6349–6361

  9. [17]

    Yu Huang, Gary G Yen, and Vincent S Tseng. 2022. Snippet policy network v2: Knee-guided neuroevolution for multi-lead ecg early classification. IEEE Transactions on Neural Networks and Learning Systems 35, 2 (2022), 2167–2181

  10. [18]

    Diederik P Kingma and Jimmy Lei Ba. 2015. Adam: A method for stochastic gradient descent. In ICLR: international conference on learning representations. ICLR US., 1–15

  11. [19]

    Achim Klenke. 2013. Probability theory: a comprehensive course. Springer Science & Business Media

  12. [20]

    Jimmy Lei Ba, Jamie Ryan Kiros, and Geoffrey E Hinton. 2016. Layer normaliza- tion. ArXiv e-prints (2016), arXiv–1607

  13. [21]

    Sergey Levine, Aviral Kumar, George Tucker, and Justin Fu. 2020. Offline rein- forcement learning: Tutorial, review, and perspectives on open problems. arXiv preprint arXiv:2005.01643 (2020)

  14. [22]

    TP Lillicrap. 2015. Continuous control with deep reinforcement learning. arXiv preprint arXiv:1509.02971 (2015)

  15. [23]

    Junwei Lv, Yuqi Chu, Jun Hu, Peipei Li, and Xuegang Hu. 2023. Second-order Confidence Network for Early Classification of Time Series. ACM Transactions on Intelligent Systems and Technology (2023)

  16. [24]

    Junwei Lv, Xuegang Hu, Lei Li, and Peipei Li. 2019. An effective confidence-based early classification of time series. IEEE Access 7 (2019), 96113–96124

  17. [25]

    Yifang Ma, Zhenyu Wang, Hong Yang, and Lin Yang. 2020. Artificial intelligence applications in the development of autonomous vehicles: a survey. IEEE/CAA Journal of Automatica Sinica 7, 2 (2020), 315–329

  18. [26]

    Coralie Martinez, Guillaume Perrin, Emmanuel Ramasso, and Michèle Rombaut

  19. [27]

    Coralie Martinez, Emmanuel Ramasso, Guillaume Perrin, and Michèle Rombaut

  20. [28]

    Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Andrei A Rusu, Joel Veness, Marc G Bellemare, Alex Graves, Martin Riedmiller, Andreas K Fidjeland, Georg Ostrovski, et al. 2015. Human-level control through deep reinforcement learning. nature 518, 7540 (2015), 529–533

  21. [29]

    Usue Mori, Alexander Mendiburu, Sanjoy Dasgupta, and Jose A Lozano. 2017. Early classification of time series by simultaneously optimizing the accuracy and earliness. IEEE transactions on neural networks and learning systems 29, 10 (2017), 4569–4578

  22. [30]

    Usue Mori, Alexander Mendiburu, Isabel Marta Miranda, and José Antonio Lozano. 2019. Early classification of time series using multi-objective optimiza- tion techniques. Information Sciences 492 (2019), 204–218

  23. [31]

    Allen Nie, Yannis Flet-Berliac, Deon Jordan, William Steenbergen, and Emma Brunskill. 2022. Data-efficient pipeline for offline reinforcement learning with limited data. Advances in Neural Information Processing Systems 35 (2022), 14810–14823

  24. [32]

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. 2019. Pytorch: An imperative style, high-performance deep learning library.Advances in neural information processing systems...

  25. [33]

    Rafael Figueiredo Prudencio, Marcos ROA Maximo, and Esther Luna Colombini

  26. [34]

    Yongyi Ran, Xin Zhou, Pengfeng Lin, Yonggang Wen, and Ruilong Deng. 2019. A survey of predictive maintenance: Systems, purposes and approaches. arXiv preprint arXiv:1912.07383 (2019)

  27. [35]

    Aurélien Renault, Youssef Achenchabe, Édouard Bertrand, Alexis Bondu, Antoine Cornuéjols, Vincent Lemaire, and Asma Dachraoui. 2024. ml_edm package: a Python toolkit for Machine Learning based Early Decision Making.arXiv e-prints (2024), arXiv–2408

  28. [36]

    Aurélien Renault, Alexis Bondu, Antoine Cornuéjols, and Vincent Lemaire. 2024. Early Classification of Time Series: Taxonomy and Benchmark. arXiv preprint arXiv:2406.18332 (2024)

  29. [37]

    Liran Ringel, Regev Cohen, Daniel Freedman, Michael Elad, and Yaniv Romano

  30. [38]

    Kauffmann, Robert A

    Lukas Ruff, Jacob R. Kauffmann, Robert A. Vandermeulen, Grégoire Montavon, Wojciech Samek, Marius Kloft, Thomas G. Dietterich, and Klaus-Robert Müller

  31. [39]

    Alejandro Pasos Ruiz, Michael Flynn, James Large, Matthew Middlehurst, and Anthony Bagnall. 2021. The great multivariate time series classification bake off: a review and experimental evaluation of recent algorithmic advances. Data Mining and Knowledge Discovery 35, 2 (2021), 401–449

  32. [40]

    IEEE Transactions on Neural Networks and Learning Systems (2023)

    A survey on offline reinforcement learning: Taxonomy, review, and open problems. IEEE Transactions on Neural Networks and Learning Systems (2023)

  33. [41]

    Patrick Schäfer and Ulf Leser. 2020. TEASER: early and accurate time series classification. Data mining and knowledge discovery 34, 5 (2020), 1336–1362

  34. [42]

    Richard S Sutton and Andrew G Barto. 2018. Reinforcement learning: An introduction. MIT press

  35. [43]

    Chang Wei Tan, Christoph Bergmeir, François Petitjean, and Geoffrey I Webb

  36. [44]

    Denis Tarasov, Vladislav Kurenkov, Alexander Nikulin, and Sergey Kolesnikov

  37. [45]

    Romain Tavenard and Simon Malinowski. 2016. Cost-aware early classification of time series. In Machine Learning and Knowledge Discovery in Databases: European Conference, ECML PKDD 2016, Riva del Garda, Italy,September 19-23, 2016, Proceedings, Part I 16. Springer, 632–647

  38. [46]

    Hado Van Hasselt, Arthur Guez, and David Silver. 2016. Deep reinforcement learning with double q-learning. In Proceedings of the AAAI conference on artificial intelligence, Vol. 30. Renault et al

  39. [47]

    A Unifying Review of Deep and Shallow Anomaly Detection. Proc. IEEE 109, 5 (2021), 756–795. https://doi.org/10.1109/JPROC.2021.3052449

  40. [48]

    Yifan Wang, Qining Zhang, Lei Ying, and Chuan Zhou. 2024. Deep Reinforce- ment Learning for Early Diagnosis of Lung Cancer. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 38. 22410–22419

  41. [49]

    Marc Rußwurm, Nicolas Courty, Rémi Emonet, Sébastien Lefèvre, Devis Tuia, and Romain Tavenard. 2023. End-to-end learned early classification of time series for in-season crop type mapping. ISPRS Journal of Photogrammetry and Remote Sensing 196 (2023), 445–456

  42. [50]

    Zhengzheng Xing, Jian Pei, and S Yu Philip. 2009. Early Prediction on Time Series: A Nearest Neighbor Approach.. In IJCAI. Citeseer, 1297–1302

  43. [51]

    Zhengzheng Xing, Jian Pei, and Philip S Yu. 2012. Early classification on time series. Knowledge and information systems 31 (2012), 105–127

  44. [52]

    Zhengzheng Xing, Jian Pei, Philip S Yu, and Ke Wang. 2011. Extracting inter- pretable features for early classification on time series. In Proceedings of the 2011 SIAM international conference on data mining. SIAM, 247–258

  45. [53]

    Data Mining and Knowledge Discovery 35, 3 (2021), 1032–1060

    Time series extrinsic regression: Predicting numeric values from time series data. Data Mining and Knowledge Discovery 35, 3 (2021), 1032–1060

  46. [55]

    Advances in Neural Information Processing Systems 36 (2024)

    Revisiting the minimalist approach to offline reinforcement learning. Advances in Neural Information Processing Systems 36 (2024)

  47. [58]

    Abraham Wald and Jacob Wolfowitz. 1948. Optimum character of the sequential probability ratio test. The Annals of Mathematical Statistics (1948), 326–339

  48. [60]

    Renjie Wu, Audrey Der, and Eamonn Keogh. 2021. When is early classification of time series meaningful. IEEE Transactions on Knowledge and Data Engineering (2021)

  49. [64]

    Paul-Emile Zafar, Youssef Achenchabe, Alexis Bondu, Antoine Cornuéjols, and Vincent Lemaire. 2021. Early classification of time series: Cost-based multi- class algorithms. In 2021 IEEE 8th International Conference on Data Science and Advanced Analytics (DSAA). IEEE, 1–10. Appe...

  50. [2017]

    Data mining and knowledge discovery 31 (2017), 606–660

    The great time series classification bake off: a review and experimental eval- uation of recent algorithmic advances. Data mining and knowledge discovery 31 (2017), 606–660

  51. [2018]

    In 2018 26th European Signal Processing Conference (EUSIPCO)

    A deep reinforcement learning approach for early classification of time series. In 2018 26th European Signal Processing Conference (EUSIPCO). IEEE, 2030–2034

  52. [2019]

    IEEE/CAA Journal of Automatica Sinica 6, 6 (2019), 1293–1305

    The UCR time series archive. IEEE/CAA Journal of Automatica Sinica 6, 6 (2019), 1293–1305

  53. [2020]

    Knowledge-Based Systems 190 (2020), 105290

    Adaptive early classification of temporal sequences using deep reinforce- ment learning. Knowledge-Based Systems 190 (2020), 105290

  54. [2021]

    Machine Learning 110, 6 (2021), 1481–1504

    Early classification of time series: Cost-based optimization criterion and algorithms. Machine Learning 110, 6 (2021), 1481–1504

  55. [2023]

    In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

    Toward Asymptotic Optimality: Sequential Unsupervised Regression of Density Ratio for Early Classification. In ICASSP 2023-2023 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 1–5

  56. [2024]

    In Forty-first International Conference on Machine Learning

    Early Time Classification with Accumulated Accuracy Gap Control. In Forty-first International Conference on Machine Learning

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.