Pith. sign in

REVIEW 2 major objections 5 minor 51 references

Evolving Markov Chains: Unsupervised Mode Discovery and Recognition from Data Streams

T0 review · 2 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Evolving Markov chains track, discover, and recognize behavioral modes online without labels.

desk verdict Solid online mode discovery with a genuine algorithmic extension, but a pseudocode bug makes the published real-world algorithm differ from the one evaluated. read the letter →

arxiv 2411.17528 v1 pith:DJMQ234V submitted 2024-11-26 cs.LG

Kutalm{i}ş Coşkun , Borahan Tümer , Bjarne C. Hiller , Martin Becker This is my paper · ORCID
classification cs.LG
keywords incrementallearningonlinemodediscoveryrecognitionnon-stationarysystemsconceptdriftMarkovchainsdatastreamschangepointdetection
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 claims that a single online update rule can maintain a $k$th-order Markov-chain model of a categorical data stream while discovering, recognizing, and switching between behavioral modes—segments of the stream generated by different stationary transition laws. The rule multiplies the matching conditional probability toward one, decays its competitors, and leaves unrelated entries untouched, so each observation costs only as much as one conditional distribution, not the whole tensor. The paper proves that the expected estimate converges to the true transition probability tensor for an ergodic $k$th-order Markov chain, and shows on synthetic and real streams—human activity, motor vibrations, EEG eye state—that the resulting mode assignments beat window- and distribution-based baselines. If the claim holds, live processes can be segmented and monitored online without ground-truth labels, known mode counts, or annotated change points.

What carries the argument

The carrying object is the $k$th-order conditional-probability tensor $\hat P[n]$, whose entries are updated by the SLWE-style rule (Eq. 2), named for the Stochastic Learning Weak Estimator on which it is based: the entry matching the observed $k+1$-gram is bumped up, the other $m-1$ entries conditioned on the same $k$-gram are multiplied by $\lambda$, and entries conditioned on other contexts are unchanged. The tensor is the mode model; the Hellinger distance between a current estimate and a $\tau$-step-old snapshot is the drift detector that triggers mode switches; a mode memory stores one tensor per discovered mode and is consulted on drift-to-steady transitions; fast and slow learning coefficients ($\lambda_f$, $\lambda_s$) adapt the update speed depending on whether the stream is in a drift or steady phase.

What would settle it

Run the estimator on a two-state first-order ergodic chain with a known transition matrix, e.g. $p(1\mid 1)=0.9$, $p(2\mid 2)=0.8$, with $\lambda=0.9$, averaging many independent runs at a large $n$. If $|\mathbb{E}[\hat p(1\mid 1)[n]] - 0.9|$ does not shrink toward zero as $n$ grows, Theorem 1 is false. Alternatively, feed EMC a stream built from two modes with identical transition tensors but different labels: the method cannot separate them by construction, so a claim that it does would indicate the evaluation protocol, not the model, is doing the work.

Watch

Extended reading notes

Core claim

The central claim is that the stochastic tensor of a $k$th-order Markov chain can be estimated online, and that drift in that tensor is itself a signal for discovering and recognizing modes. The update, for each observed symbol, treats the current context $\langle s_1,\dots,s_k\rangle$: the estimated probability of the observed next symbol is moved toward $1$ by adding $(1-\lambda)$, every alternative next symbol is scaled by $\lambda$, and all entries belonging to other contexts are left alone. Theorem 1 states that for an ergodic $k$th-order chain this estimator is weakly convergent, $\mathbb{E}(\hat P[\infty]) = P$, and because the recurrence is multiplicative, convergence is geometric. The estimated tensor is compared against a delayed snapshot using Hellinger distance; a rise above a threshold flags a drift, a return below it ends the drift, and a memory of previously stored mode tensors lets the algorithm either recognize a recurring mode or create a new one. Entropy regulation and fast/slow learning are auxiliary mechanisms that keep rarely seen contexts from freezing stale probabilities and that trade estimation variance against adaptation speed.

Load-bearing premise

The load-bearing premise is that each mode is a stationary $k$th-order ergodic Markov chain over a fixed, known alphabet $\Sigma$ with a fixed order $k$; the convergence theorem does not by itself cover non-stationary streams, where tracking is justified only by the empirical results.

Editorial extensions

If this is right

  • A stream of categorical observations can be segmented into behavioral regimes in real time with no labels, no known mode count, and no annotated change points.
  • Per-observation cost is $O(m^k)$ rather than $O(m^{k+1})$, because only the one active conditional distribution is updated, so higher-order dependencies remain tractable for moderate alphabets.
  • Expected transition estimates converge to the true tensor within each stationary regime, so longer regimes mean more accurate per-mode models in memory.
  • Recurring modes are recognized and their stored models are refined incrementally whenever the stream is steady, which supports non-stationary processes whose behaviors repeat.
  • The same machinery handles probability tracking, change-point detection, mode discovery, and mode recognition within one framework, as shown by a synthetic change-point F1 of 0.93 and a synthetic mode-discovery ARI of 0.85.

Reading between the lines

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

  • The proof of convergence treats each mode as a fixed ergodic chain; a formal analysis of the same update under slowly time-varying transition matrices would quantify the tracking lag, which the paper currently assesses only empirically.
  • Entropy regulation is a second forgetting mechanism acting on the whole tensor; its interaction with the multiplicative $\lambda$-forgetting is not characterized theoretically, so a combined analysis could predict when the two work against each other.
  • The fixed-alphabet and fixed-order assumptions could be relaxed by letting rarely used states split or merge; the paper lists this as future work, and it would turn EMC from a mode tracker into an adaptively structured model.
  • The Hellinger-distance drift detector compares whole tensors; replacing it with context-specific distances could yield finer-grained drift localization, identifying which transition probabilities changed, at the cost of more memory.
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 / 5 minor

Summary. The paper proposes Evolving Markov Chains (EMCs), an online, unsupervised method for mode discovery and recognition from categorical data streams. EMCs maintain a kth-order Markov-chain probability tensor updated with a multiplicative rule derived from Stochastic Learning Weak Estimation (SLWE), detect drift via periodic Hellinger-distance comparisons, and store per-mode tensors in memory to recognize recurring regimes. The authors claim an O(m^k) per-observation update, convergence of the expected estimates to the true transition probabilities, and demonstrate the method on synthetic data, human activity recognition, electric motor condition monitoring, and EEG eye-state data.

Significance. If the claims hold, the paper offers a useful and efficient primitive for online temporal clustering and change-point detection: it avoids tracking windows, updates only the relevant tensor slice, handles arbitrary Markov order, and comes with released code and a broad empirical comparison against stream-clustering and change-detection baselines. The formal statement, however, only covers convergence in expectation for a fixed ergodic chain, and the non-stationary mode-switching behavior is supported empirically. The main strength is the combination of a simple, efficient update rule with a practical mode-memory mechanism; the main weakness is a reproducibility-relevant mismatch between the mathematical specification of entropy regulation and the pseudocode used for the real-world experiments.

major comments (2)
  1. The pseudocode does not implement the entropy regulation defined in Eq. (9). Eq. (9) requires Q[n] = (1−β)Q[n−1] + βU with U fixed as the uniform distribution. In Algorithm 1, line 8 first overwrites U(·|c) with the just-updated P^P(·|c), and line 9 then computes P^← (1−β)P^+ βU. For the CPD at context c this gives (1−β)P^+ βP^= P^, so no entropy regulation is applied; if U is instead read as the current CPD, the block is a no-op. Since all real-world experiments use β > 0 (β = 0.01, 0.003, 0.001 in Supplementary Table A4 for HAR, CWRU, and EEG respectively), the reported real-world results are produced by an algorithm different from the one specified in the manuscript. This is a load-bearing reproducibility issue: please correct Algorithm 1 (or Eq. (9)) and confirm with the released code that the evaluated algorithm matches the corrected specification.
  2. The proof establishes only convergence of E(P^[n]) to the true probability; it does not provide variance bounds, almost-sure convergence, or the 'geometric convergence' claimed in Section 5.1 and the abstract. Since the mode-switching and drift-detection behavior relies on the estimate being close to the true tensor, a mean-only statement is weak support for the adaptation claim. Please either provide a concentration or variance result, or explicitly limit Theorem 1 and the abstract's convergence claim to convergence in expectation.
minor comments (5)
  1. The formula for the Hellinger distance is typeset ambiguously; it should read H(P,Q) = (1/√2) * sqrt(Σ_i (√p_i − √q_i)²).
  2. Lines 3 and 14 contain 'append M P to M' / 'append M P to M', which appears to be a typo for appending the estimated tensor P^ to the mode memory M; please correct the notation.
  3. The text says the first 3000 instances constitute 20% of the EEG sequence, but the total sequence length is not stated; please give the total length for clarity.
  4. The caption states loads from 0 to 4 HP, but the table only reports 0, 1, 2, and 3 HP; please align the caption with the data.
  5. Theorem 1 is described as 'weak convergence', but the statement E(P^[∞]) = P is convergence in expectation, not weak convergence in the usual probabilistic sense; please adjust the terminology to avoid confusion.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the estimator derivation and Theorem 1 are self-contained, and real-world evaluations use external labels; self-citations are baselines and preprocessing, not load-bearing premises.

full rationale

The central derivation is not circular. Eq. (2) defines the online update of the kth-order conditional probability estimates, and Theorem 1 proves E(hat P[∞]) = P directly from the update's conditional expectation (Eqs. (3)-(8)); it does not assume its own conclusion or import a key theorem from the authors' prior work. The cited geometric-convergence statements [32,33] are external and used only to justify speed of convergence. The self-citations are not load-bearing: [27] (SCD) is a comparison baseline and a motivation, [23] is used only for primitive-alphabet preprocessing in real-world experiments, [35] adds fast/slow learning, and [22] is an offline baseline; none is invoked to forbid alternatives or to supply the uniqueness of the method. Synthetic experiments are in-distribution (data generated from the same kth-order Markov-chain model class that EMC assumes), but they are ground-truth evaluations, not reductions of the reported ARI to a fitted value, and the real-world HAR, CWRU, and EEG results are measured against external labels after standard hyperparameter tuning on hold-out portions. One reproducibility concern, not a circularity, is that Algorithm 1 lines 7-11 do not literally implement Eq. (9) for the just-updated CPD: line 8 sets U(·|c) to the current hat P(·|c), so the subsequent convex combination leaves that CPD unchanged; this is an implementation/specification inconsistency that should be resolved by the released code, but it does not make any prediction equivalent to its input by construction.

Assumptions & free parameters 7 free parameters · 3 assumptions · 0 invented entities

The method depends on several tuned hyperparameters and on the modeling assumption that each mode is a fixed-order Markov chain over a fixed alphabet. The theoretical result requires ergodicity within a mode, which is at odds with the non-stationary setting the algorithm targets. No new physical or latent entities are introduced; 'modes' are a standard behavioral-state concept.

free parameters (7)
  • lambda_f (fast learning coefficient) = 0.91 to 0.94 in experiments
    Controls update magnitude during drift; tuned via Bayesian optimization on validation data.
  • lambda_s (slow learning coefficient) = 0.95 to 0.97 in experiments
    Controls update magnitude during steady conditions; tuned via Bayesian optimization.
  • beta (entropy regularization rate) = 0 to 0.01 in experiments
    Controls rate at which unused conditional probabilities drift toward uniform; pseudocode inconsistency makes its effective value unclear.
  • delta (drift threshold) = 0.05 to 0.4 in experiments
    Threshold on Hellinger distance between tensors to declare drift; tuned per experiment.
  • eta (mode similarity threshold) = 0.07 to 0.5 in experiments
    Threshold for matching current estimate to a stored mode; tuned per experiment.
  • tau (drift check interval) = 25 to 100 in experiments
    Number of steps between Hellinger comparisons; set relative to 1/(1-lambda).
  • Markov order k = 1 or 2
    Chosen per application; determines tensor size and update cost; not learned from data.
assumptions (3)
  • domain assumption Each mode is a kth-order Markov chain over a fixed alphabet Σ.
    Assumed in Section 5.1 (Eq. 2) and used by the synthetic generator; if a mode's state space or order differs, the tensor is mis-specified.
  • domain assumption Within a mode, the process is ergodic and stationary for the convergence theorem.
    Theorem 1 in Section 5.1 requires ergodicity to treat the history probability A as constant; non-stationary tracking is not covered by the proof.
  • standard math Geometric convergence of the SLWE update carries over to the conditional probability extension.
    The paper states this and cites [32, 33]; it does not prove it for Eq. (2).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Evolving Markov Chains: Unsupervised Mode Discovery and Recognition from Data Streams." pith.science (2026). https://pith.science/paper/DJMQ234V

@misc{pith2026241117528,
  author       = {Pith},
  title        = {Pith review of: Evolving Markov Chains: Unsupervised Mode Discovery and Recognition from Data Streams},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DJMQ234V}},
  note         = {Machine review of arXiv:2411.17528}
}
read the original abstract

Markov chains are simple yet powerful mathematical structures to model temporally dependent processes. They generally assume stationary data, i.e., fixed transition probabilities between observations/states. However, live, real-world processes, like in the context of activity tracking, biological time series, or industrial monitoring, often switch behavior over time. Such behavior switches can be modeled as transitions between higher-level \emph{modes} (e.g., running, walking, etc.). Yet all modes are usually not previously known, often exhibit vastly differing transition probabilities, and can switch unpredictably. Thus, to track behavior changes of live, real-world processes, this study proposes an online and efficient method to construct Evolving Markov chains (EMCs). EMCs adaptively track transition probabilities, automatically discover modes, and detect mode switches in an online manner. In contrast to previous work, EMCs are of arbitrary order, the proposed update scheme does not rely on tracking windows, only updates the relevant region of the probability tensor, and enjoys geometric convergence of the expected estimates. Our evaluation of synthetic data and real-world applications on human activity recognition, electric motor condition monitoring, and eye-state recognition from electroencephalography (EEG) measurements illustrates the versatility of the approach and points to the potential of EMCs to efficiently track, model, and understand live, real-world processes.

Figures

Figures reproduced from arXiv: 2411.17528 by the authors.

Figure 1
Figure 1. (a) The black box process has unknown underlying modes ( [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Method Regime Duration U(500, 1000) U(1500, 2000) U(2500, 3000) EMC 0.056 ± 0.007 0.036 ± 0.005 0.029 ± 0.004 MC-ADWIN 0.086 ± 0.012 0.063 ± 0.014 0.054 ± 0.015 MC-SW (w = 100) 0.075 ± 0.004 0.071 ± 0.004 0.070 ± 0.004 MC-SW (w = 328) 0.068 ± 0.005 0.050 ± 0.003 0.045 ± 0.003 MC-SW (w = 500) 0.080 ± 0.007 0.051 ± 0.003 0.043 ± 0.003 [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 3
Figure 3. Transition plots of true modes (top) and detected modes (bottom) on synthetic data. The heatmap in [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Transition plots of true modes (top) and detected modes (bottom) on activity data of subject [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Result on vibration data under 2 HP of load with mode transition sequence of OK → FD:07 → FD:14. The labels of each condition is given along the top horizontal axis. The heatmap in the bottom plot represents the Hellinger distance, quantifying the detected drift over t…
Figure 6
Figure 6. Figure 6: Results of EEG eye-state task. ARI becomes [PITH_FULL_IMAGE:figures/full_fig_p014_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

51 extracted references · 34 canonical work pages

  1. [1]

    Big Data Analytics in Healthcare: A Systematic Literature Review

    Sayantan Khanra et al. “Big Data Analytics in Healthcare: A Systematic Literature Review”. In:Enterprise Information Systems14.7 (Aug. 2020), pp. 878–912.issn: 1751-7575, 1751-7583.doi: 10.1080/17517575. 2020.1812005

    work page arXiv 2020

  2. [2]

    Big Data Analytics for Intelligent Manufacturing Systems: A Review

    Junliang Wang et al. “Big Data Analytics for Intelligent Manufacturing Systems: A Review”. In:Journal of Manufacturing Systems 62 (Jan. 2022), pp. 738–752.issn: 02786125. doi: 10.1016/j.jmsy.2021.03.005

    doi:10.1016/j.jmsy.2021.03.005 2022

  3. [3]

    Buede.The Engineering Design of Systems: Models and Methods

    Dennis M. Buede.The Engineering Design of Systems: Models and Methods. 1st ed. Wiley, Jan. 2009.isbn: 978-0-470-16402-0. doi: 10.1002/9780470413791

    work page doi:10.1002/9780470413791 2009

  4. [4]

    An Architectural Framework of Elderly Healthcare Monitoring and Tracking through Wearable Sensor Technologies

    Abeer Alsadoon, Ghazi Al-Naymat, and Oday D. Jerew. “An Architectural Framework of Elderly Healthcare Monitoring and Tracking through Wearable Sensor Technologies”. In:Multimedia Tools and Applications 83.26 (Jan. 2024), pp. 67825–67870.issn: 1573-7721. doi: 10.1007/s11042-024-18177-0

    work page doi:10.1007/s11042-024-18177-0 2024

  5. [5]

    Anomaly Monitoring Improves Remaining Useful Life Estimation of Industrial Machinery

    Gurkan Aydemir and Burak Acar. “Anomaly Monitoring Improves Remaining Useful Life Estimation of Industrial Machinery”. In:Journal of Manufacturing Systems56 (July 2020), pp. 463–469.issn: 02786125. doi: 10.1016/j.jmsy.2020.06.014. 15

    work page doi:10.1016/j.jmsy.2020.06.014 2020

  6. [6]

    Data Stream Clustering: A Review

    Alaettin Zubaroğlu and Volkan Atalay. “Data Stream Clustering: A Review”. In:Artificial Intelligence Review 54.2 (Feb. 2021), pp. 1201–1236.issn: 0269-2821, 1573-7462.doi: 10.1007/s10462-020-09874-x

    work page doi:10.1007/s10462-020-09874-x 2021

  7. [7]

    Data Stream Clustering: A Survey

    Jonathan A. Silva et al. “Data Stream Clustering: A Survey”. In:ACM Computing Surveys46.1 (Oct. 2013), pp. 1–31. issn: 0360-0300, 1557-7341.doi: 10.1145/2522968.2522981

    work page arXiv 2013

  8. [8]

    A Survey of Stream Clustering Algorithms

    Charu C. Aggarwal. “A Survey of Stream Clustering Algorithms”. In:Data Clustering. Ed. by Charu C. Aggarwal and Chandan K. Reddy. 1st ed. Chapman and Hall/CRC, Sept. 2018, pp. 231–258.isbn: 978-1-315-37351-5. doi: 10.1201/9781315373515-10

    work page doi:10.1201/9781315373515-10 2018

Show all 51 references
  1. [9]

    DataStreamClusteringAlgorithms: A Review

    MaryamMousavi,AzuralizaAbuBakar,andMohammadmahdiVakilian.“DataStreamClusteringAlgorithms: A Review”. In:Int J Adv Soft Comput Appl7.3 (2015), p. 13

    2015

  2. [10]

    StreamKM++: A Clustering Algorithm for Data Streams

    Marcel R. Ackermann et al. “StreamKM++: A Clustering Algorithm for Data Streams”. In:ACM Journal of Experimental Algorithmics17 (July 2012).issn: 1084-6654, 1084-6654.doi: 10.1145/2133803.2184450

    2012

  3. [11]

    Improved Clustering Algorithm Based on High-Speed Network Data Stream

    Chunyong Yin et al. “Improved Clustering Algorithm Based on High-Speed Network Data Stream”. In:Soft Computing 22.13 (July 2018), pp. 4185–4195.issn: 1432-7643, 1433-7479.doi: 10.1007/s00500-017-2708-2

    2018 doi

  4. [12]

    Adaptive Clustering for Dynamic IoT Data Streams

    Daniel Puschmann, Payam Barnaghi, and Rahim Tafazolli. “Adaptive Clustering for Dynamic IoT Data Streams”. In:IEEE Internet of Things Journal4.1 (Feb. 2017), pp. 64–74.issn: 2327-4662, 2372-2541.doi: 10.1109/JIOT.2016.2618909

    2017

  5. [13]

    An Evolutionary Algorithm for Clustering Data Streams with a Variable Number of Clusters

    Jonathan De Andrade Silva, Eduardo Raul Hruschka, and João Gama. “An Evolutionary Algorithm for Clustering Data Streams with a Variable Number of Clusters”. In:Expert Systems with Applications67 (Jan. 2017), pp. 228–238.issn: 09574174. doi: 10.1016/j.eswa.2016.09.020

    2017 doi

  6. [14]

    Fully Online Clustering of Evolving Data Streams into Arbitrarily Shaped Clusters

    Richard Hyde, Plamen Angelov, and A.R. MacKenzie. “Fully Online Clustering of Evolving Data Streams into Arbitrarily Shaped Clusters”. In:Information Sciences382–383 (Mar. 2017), pp. 96–114.issn: 00200255. doi: 10.1016/j.ins.2016.12.004

    2017 doi

  7. [15]

    DBIECM-an Evolving Clustering Method for Streaming Data Clustering

    Kaisong Zhang et al. “DBIECM-an Evolving Clustering Method for Streaming Data Clustering”. In:Advances in Modelling and Analysis B60.1 (Mar. 2017), pp. 239–254.issn: 12404543. doi: 10.18280/ama_b.600115

    2017 doi

  8. [16]

    On Clustering Massive Text and Categorical Data Streams

    Charu C. Aggarwal and Philip S. Yu. “On Clustering Massive Text and Categorical Data Streams”. In: Knowledge and Information Systems 24.2 (Aug. 2010), pp. 171–196. issn: 0219-1377, 0219-3116. doi: 10.1007/s10115-009-0241-z

    2010 doi

  9. [17]

    Wai-Ki Ching et al.Markov Chains: Models, Algorithms and Applications. Vol. 189. International Series in Operations Research & Management Science. Boston, MA: Springer US, 2013.isbn: 978-1-4614-6311-5. doi: 10.1007/978-1-4614-6312-2

    2013 doi

  10. [18]

    Hoboken, NJ, USA: John Wiley & Sons, Inc., June 2017.isbn: 978-1-119-38759-6

    Paul A Gagniuc.Markov Chains: From Theory to Implementation and Experimentation. Hoboken, NJ, USA: John Wiley & Sons, Inc., June 2017.isbn: 978-1-119-38759-6. doi: 10.1002/9781119387596

    2017 doi

  11. [19]

    Evolutionary Markov Dynamics for Network Community Detection

    Zhen Wang et al. “Evolutionary Markov Dynamics for Network Community Detection”. In:IEEE Transactions on Knowledge and Data Engineering34.3 (Mar. 2022), pp. 1206–1220.issn: 1041-4347, 1558-2191, 2326-3865. doi: 10.1109/TKDE.2020.2997043

    2022

  12. [20]

    Improving the Estimation of Markov Transition Probabilities Using Mechanistic- Empirical Models

    Daijiro Mizutani et al. “Improving the Estimation of Markov Transition Probabilities Using Mechanistic- Empirical Models”. In:Frontiers in Built Environment3 (Oct. 2017), p. 58.issn: 2297-3362. doi: 10.3389/ fbuil.2017.00058

    2017

  13. [21]

    Fast and Adaptive Variable Order Markov Chain Construction

    Marcel H. Schulz et al. “Fast and Adaptive Variable Order Markov Chain Construction”. In:Algorithms in Bioinformatics. Ed. by Keith A. Crandall and Jens Lagergren. Vol. 5251. Berlin, Heidelberg: Springer Berlin Heidelberg, 2008, pp. 306–317.isbn: 978-3-540-87360-0. doi: 10.100...

    2008 doi

  14. [22]

    Multivariate Time Series Clustering and Its Application in Industrial Systems

    Barış Gün Sürmeli and M. Borahan Tümer. “Multivariate Time Series Clustering and Its Application in Industrial Systems”. In:Cybernetics and Systems51.3 (Apr. 2020), pp. 315–334.issn: 0196-9722, 1087-6553. doi: 10.1080/01969722.2019.1691851

    2020

  15. [23]

    A Syntactic Pattern Recognition Based Approach to Online Anomaly Detection and Identification on Electric Motors

    Kutalmış Coşkun et al. “A Syntactic Pattern Recognition Based Approach to Online Anomaly Detection and Identification on Electric Motors”. In:Pattern Recognition. Ed. by Björn Andres et al. Vol. 13485. Cham: Springer International Publishing, 2022, pp. 116–132.isbn: 978-3-031-...

    2022 doi

  16. [24]

    Anomaly Detection Based on a Dynamic Markov Model

    Huorong Ren, Zhixing Ye, and Zhiwu Li. “Anomaly Detection Based on a Dynamic Markov Model”. In: Information Sciences411 (Oct. 2017), pp. 52–65.issn: 00200255. doi: 10.1016/j.ins.2017.05.021

    2017 doi

  17. [25]

    Cooperative Multiagent System for Park- ing Availability Prediction Based on Time Varying Dynamic Markov Chains

    Surafel Luleseged Tilahun and Giovanna Di Marzo Serugendo. “Cooperative Multiagent System for Park- ing Availability Prediction Based on Time Varying Dynamic Markov Chains”. In:Journal of Advanced Transportation2017 (2017), pp. 1–14.issn: 0197-6729, 2042-3195.doi: 10.1155/2017/1760842

    2017 doi

  18. [26]

    Stochastic Optimal Control of Systems with Soft Constraints and Opportunities for Automotive Applications

    I. V. Kolmanovsky and D. P. Filev. “Stochastic Optimal Control of Systems with Soft Constraints and Opportunities for Automotive Applications”. In:2009 IEEE International Conference on Control Applications. St. Petersburg, Russia: IEEE, July 2009, pp. 1265–1270.isbn: 978-1-424...

    2009 doi

  19. [27]

    Detection of Regime Switching Points in Non-Stationary Sequences Using Stochastic Learning Based Weak Estimation Method

    Ezdin Aslanci et al. “Detection of Regime Switching Points in Non-Stationary Sequences Using Stochastic Learning Based Weak Estimation Method”. In:2017 IEEE 15th International Conference on Industrial Informatics (INDIN). Emden: IEEE, July 2017, pp. 787–792.isbn: 978-1-5386-08...

    2017

  20. [28]

    Temporal Structure Learning for Clustering Massive Data Streams in Real-Time

    Michael Hahsler and Margaret H. Dunham. “Temporal Structure Learning for Clustering Massive Data Streams in Real-Time”. In: Proceedings of the 2011 SIAM International Conference on Data Mining. Society for Industrial and Applied Mathematics, Apr. 2011, pp. 664–675.isbn: 978-0-...

    2011 doi

  21. [29]

    George E. P. Box et al.Time Series Analysis: Forecasting and Control. Fifth. Wiley Series in Probability and Statistics. Hoboken, New Jersey: John Wiley & Sons, Inc, 2016.isbn: 978-1-118-67502-1

    2016

  22. [30]

    Learning in Nonstationary Environments: A Survey

    Gregory Ditzler et al. “Learning in Nonstationary Environments: A Survey”. In:IEEE Computational Intelligence Magazine 10.4 (Nov. 2015), pp. 12–25.issn: 1556-603X. doi: 10.1109/MCI.2015.2471196

    2015

  23. [31]

    Learning under Concept Drift: A Review

    Jie Lu et al. “Learning under Concept Drift: A Review”. In:IEEE Transactions on Knowledge and Data Engineering (2018), pp. 1–1.issn: 1041-4347, 1558-2191, 2326-3865.doi: 10.1109/TKDE.2018.2876857

    2018

  24. [32]

    Stochastic Learning-Based Weak Estimation of Multinomial Random Variables and Its Applications to Pattern Recognition in Non-Stationary Environments

    B. John Oommen and Luis Rueda. “Stochastic Learning-Based Weak Estimation of Multinomial Random Variables and Its Applications to Pattern Recognition in Non-Stationary Environments”. In: Pattern Recognition39.3 (Mar. 2006), pp. 328–341.issn: 00313203. doi: 10.1016/j.patcog.2005.09.007

    2006 doi

  25. [33]

    Narendra and Mandayam A

    Kumpati S. Narendra and Mandayam A. L. Thathachar.Learning Automata: An Introduction. Mineola, New York: Dover Publications, Inc, 2012.isbn: 978-0-486-49877-5

    2012

  26. [34]

    Neue Begründung Der Theorie Quadratischer Formen von Unendlichvielen Veränderlichen

    E. Hellinger. “Neue Begründung Der Theorie Quadratischer Formen von Unendlichvielen Veränderlichen.” In: Journal für die reine und angewandte Mathematik1909.136 (July 1909), pp. 210–271.issn: 1435-5345, 0075-4102. doi: 10.1515/crll.1909.136.210

    1909 doi

  27. [35]

    An Adaptive Estimation Method with Exploration and Exploitation Modes for Non-Stationary Environments

    Kutalmış Coşkun and Borahan Tümer. “An Adaptive Estimation Method with Exploration and Exploitation Modes for Non-Stationary Environments”. In:Pattern Recognition129 (Sept. 2022), p. 108702.issn: 00313203. doi: 10.1016/j.patcog.2022.108702

    2022

  28. [36]

    Optuna: A Next-generation Hyperparameter Optimization Framework

    Takuya Akiba et al. “Optuna: A Next-generation Hyperparameter Optimization Framework”. In:Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 2019

    2019

  29. [37]

    Learning from Time-Changing Data with Adaptive Windowing

    Albert Bifet and Ricard Gavaldà. “Learning from Time-Changing Data with Adaptive Windowing”. In: Proceedings of the 2007 SIAM International Conference on Data Mining. Society for Industrial and Applied Mathematics, Apr. 2007, pp. 443–448.isbn: 978-0-89871-630-6. doi: 10.1137/1...

    2007 doi

  30. [38]

    Gerrit J. J. van den Burg and Christopher K. I. Williams.An Evaluation of Change Point Detection Algorithms. 2020. arXiv:2003.06222 [cs, stat] . url: http://arxiv.org/abs/2003.06222

    2020 arXiv

  31. [39]

    Reactive Soft Prototype Computing for Concept Drift Streams

    Christoph Raab, Moritz Heusinger, and Frank-Michael Schleif. “Reactive Soft Prototype Computing for Concept Drift Streams”. In:Neurocomputing 416 (Nov. 2020), pp. 340–351.issn: 09252312. doi: 10.1016/j. neucom.2019.11.111

    2020 doi

  32. [40]

    Continuous Inspection Schemes

    Ewan S Page. “Continuous Inspection Schemes”. In:Biometrika 41 (1954), p. 17

    1954

  33. [41]

    On Evaluating Stream Learning Algorithms

    João Gama, Raquel Sebastião, and Pedro Pereira Rodrigues. “On Evaluating Stream Learning Algorithms”. In: Machine Learning90.3 (Mar. 2013), pp. 317–346.issn: 0885-6125, 1573-0565.doi: 10.1007/s10994- 012-5320-9

    2013 doi

  34. [42]

    A Framework for Clustering Evolving Data Streams

    Charu C. Aggarwal et al. “A Framework for Clustering Evolving Data Streams”. In:Proceedings 2003 VLDB Conference. Elsevier, 2003, pp. 81–92.isbn: 978-0-12-722442-8. doi: 10.1016/B978-012722442-8/50016-1

    2003 doi

  35. [43]

    Clustering Data Streams Based on Shared Density between Micro- Clusters

    Michael Hahsler and Matthew Bolaos. “Clustering Data Streams Based on Shared Density between Micro- Clusters”. In:IEEE Transactions on Knowledge and Data Engineering28.6 (June 2016), pp. 1449–1461. issn: 1041-4347. doi: 10.1109/TKDE.2016.2522412

    2016

  36. [44]

    On-Body Localization of Wearable Devices: An Investigation of Position-Aware Activity Recognition

    Timo Sztyler and Heiner Stuckenschmidt. “On-Body Localization of Wearable Devices: An Investigation of Position-Aware Activity Recognition”. In:2016 IEEE International Conference on Pervasive Computing and Communications (PerCom). Sydney, Australia: IEEE, Mar. 2016, pp. 1–9.is...

    2016

  37. [45]

    King Sun Fu, Wolf Dieter Keidel, and Hans Wolter, eds.Syntactic Pattern Recognition, Applications. Vol. 14. Communication and Cybernetics. Berlin, Heidelberg: Springer Berlin Heidelberg, 1977.isbn: 978-3-642- 66440-3. doi: 10.1007/978-3-642-66438-0

    1977 doi

  38. [46]

    Bernd Fritzke.Breathing K-Means. Oct. 2021. arXiv:2006.15666 [cs, stat] . url: http://arxiv.org/ abs/2006.15666

    2021 arXiv

  39. [47]

    A Cluster Separation Measure

    David L. Davies and Donald W. Bouldin. “A Cluster Separation Measure”. In:IEEE Transactions on Pattern Analysis and Machine IntelligencePAMI-1.2 (Apr. 1979), pp. 224–227.issn: 0162-8828, 2160-9292.doi: 10.1109/TPAMI.1979.4766909. 17

    1979

  40. [48]

    Bearing Data Center | Case School of Engineering | Case Western Reserve University. Aug. 2021. url: https://engineering.case.edu/bearingdatacenter

    2021

  41. [49]

    UCI Machine Learning Repository

    Oliver Roesler.EEG Eye State. UCI Machine Learning Repository. DOI: https://doi.org/10.24432/C57G7J. 2013

    2013 doi

  42. [50]

    Segmentation of Brain Electrical Activity into Microstates: Model Estimation and Validation

    R.D. Pascual-Marqui, C.M. Michel, and D. Lehmann. “Segmentation of Brain Electrical Activity into Microstates: Model Estimation and Validation”. In:IEEE Transactions on Biomedical Engineering42.7 (July 1995), pp. 658–665.issn: 0018-9294, 1558-2531.doi: 10.1109/10.391164

    1995 doi

  43. [51]

    Pycrostates: a Python library to study EEG microstates

    Victor Férat et al. “Pycrostates: a Python library to study EEG microstates”. In:Journal of Open Source Software 7.78 (2022), p. 4564.doi: 10.21105/joss.04564. url: https://doi.org/10.21105/joss.04564. 18 A Supplemental Material A.1 Supplemental Figures 1 2 3 4 5 6 7 8 9 10 Ma...

    2022 doi

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.