Pith. sign in

REVIEW 4 major objections 5 minor 58 references

Clustering Time Series Data with Gaussian Mixture Embeddings in a Graph Autoencoder Framework

T0 review · 4 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 A graph autoencoder trained with a Gaussian-mixture latent prior achieves the best average NMI and RI among ten baselines on 19 time series datasets.

desk verdict Reasonable composition of known pieces, but the SOTA claim rests on a hyperparameter selection protocol that leaks labels; the method is worth refereeing, not the numbers as reported. read the letter →

arxiv 2411.16972 v1 pith:MKCUXNLD submitted 2024-11-25 cs.LG cs.AIeess.SP

Amirabbas Afzali , Hesam Hosseini , Mohmmadamin Mirzai , Arash Amini This is my paper · ORCID
classification cs.LGcs.AIeess.SP
keywords timeseriesclusteringgraphautoencoderGaussianmixtureembeddingsvariationalweighteddynamicwarpingunsupervisedrepresentationlearningUCRbenchmarkstockmarket
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

The paper proposes VMGAE, a graph autoencoder for unsupervised time series clustering. Each time series becomes a node in a graph whose edges come from weighted dynamic time warping distances, thresholded so that a fixed fraction of possible edges is kept. A variational graph autoencoder is trained with a loss that reconstructs the graph while regularizing the latent space into a mixture of Gaussians, and the final clusters come from a Gaussian mixture model fit to the learned embeddings. On 19 UCR datasets, the paper reports the best average NMI (0.6553) and RI (0.8605) and the lowest average rank among ten baselines, and it illustrates the method on stock-market data.

What carries the argument

The central object is the VMGAE objective, an ELBO (evidence lower bound) whose prior over the latent variable is a mixture of Gaussians rather than a single standard normal. The reconstruction term is the standard graph autoencoder loss over the adjacency matrix, and the regularizer couples each embedding to a cluster-specific Gaussian with parameters {µ̃_c, σ̃_c, π_c}, pulling the latent space into K separable components. This is supported by a graph-construction rule that fixes the edge density α and derives the WDTW distance threshold δ per dataset, avoiding a single global threshold. This combination is what the paper claims makes both the graph topology and the embedding space adaptive across heterogeneous time series.

What would settle it

Re-run VMGAE on all 19 UCR datasets with hyperparameters selected by a fully automated, label-free rule, and recompute the averages; if the label-free run no longer records the lowest average rank on both NMI and RI, the claimed state-of-the-art result depends on label-influenced model selection rather than on the method alone.

Watch

Extended reading notes

Core claim

The paper's central claim is that structuring time series as graphs before embedding them makes clustering more accurate, and that a Gaussian-mixture regularizer on the latent space is what lets a graph autoencoder separate the clusters. Concretely, VMGAE takes the WDTW distance matrix, converts it to an adjacency matrix by fixing a graph density α and solving for the distance threshold δ, then runs a two-layer GCN encoder whose latent embeddings are reparameterized as in a variational autoencoder. The decoder is an inner product that reconstructs the adjacency matrix, and the ELBO adds a regularizer that pushes the latent representation toward a mixture of K Gaussians, with responsibilities computed as p(c_i|z_i). Clusters are assigned by refitting a GMM on the final embeddings. Against ten baselines on 19 UCR datasets, the paper reports average rank 3.1579 for NMI and 2.6842 for RI, average NMI 0.6553 and RI 0.8605, and the best score on seven datasets by each metric; on three datasets (TwoPatterns, SonyAIBORobotSurface1, TwoLeadECG) the reported NMI gains over the prior best are large.

Load-bearing premise

The load-bearing premise is that the hyperparameters γ, λ, W, and α were tuned per dataset without using ground-truth labels, and that the baseline scores quoted from earlier papers were produced under a comparable protocol.

Editorial extensions

If this is right

  • Time series clustering can be treated as graph node embedding, so structural relations between series in DTW space are available to the encoder rather than only raw values.
  • The Gaussian-mixture regularizer gives each cluster an explicit mean and covariance in latent space, making the embedding interpretable as a generative model and enabling soft cluster assignments.
  • The threshold construction with fixed density α adapts the graph to each dataset without requiring a hand-chosen distance threshold δ.
  • On the 50-stock application, the procedure surfaces distinct community patterns in normalized closing prices, which the paper connects to market prediction, portfolio construction, and risk management.

Reading between the lines

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

  • A natural extension is to formalize the visual hyperparameter heuristic as an automated, label-free rule, for example a silhouette score computed on the reconstructed graph, which would make the reported per-dataset choices reproducible without ground truth.
  • An ablation that clusters directly with the learned responsibilities q(c_i|X,A), skipping the final GMM refit, would isolate how much of the gain comes from the mixture regularizer itself; the paper only states empirically that the refit improves performance.
  • Because the encoder and ELBO are graph-agnostic, VMGAE could be applied to other node-level graph clustering benchmarks beyond the single Cora example, such as social or biological networks.
  • The fixed-density construction could be paired with other elastic time series distances, such as shapeDTW, to test whether fixing α rather than δ stabilizes graph topology across distance measures.
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

4 major / 5 minor

Summary. The paper proposes VMGAE, a variational graph autoencoder for time series clustering. Time series are converted into a graph using Weighted Dynamic Time Warping (WDTW) with a window constraint, and the adjacency matrix is built by fixing a graph edge density alpha rather than an absolute distance threshold. The encoder is a two-layer GCN producing Gaussian latent embeddings, and the loss combines graph reconstruction with a mixture-of-Gaussians regularizer in the spirit of VaDE. Clustering is obtained by refitting a GMM on the learned latent representations. The paper reports NMI and RI on 19 UCR datasets, claims state-of-the-art performance, includes a qualitative financial-data case study, and adds ablations on hyperparameters and convolution layers.

Significance. If the empirical claims were fully supported, VMGAE would be a practically useful contribution: the density-based graph construction is a reasonable design idea, the ELBO derivation in Eqs. (12)-(19) is standard and appears formally sound, and the paper includes a useful sensitivity analysis and a Cora experiment showing flexibility beyond time series. However, the headline claim of significantly outperforming state-of-the-art time series clustering is not currently supported. The per-dataset hyperparameter selection uses latent visualizations colored by ground-truth labels, and Appendix F.1 documents metric-based exploration of the same grid, so the reported numbers are not a label-blind unsupervised benchmark. The baseline numbers are also largely borrowed from original papers under uncontrolled protocols, and no variance or statistical testing is reported. The contribution is therefore conditional on a substantially reworked experimental protocol.

major comments (4)
  1. [Experimental Setup; Appendix E; Appendix F.1] The state-of-the-art comparison is invalidated by label-dependent hyperparameter selection. The Experimental Setup states that gamma, lambda, W, and alpha were tuned per dataset by 'visualizing the latent distribution of the training set,' and Figure 4, which is the visualization used for that purpose, colors all points by their ground-truth class labels. Appendix F.1 then explicitly reports NMI/RI values computed across gamma values and notes that for SonyAIBORobotSurface1, gamma=0.7 and gamma=1.0 give better metrics than the selected gamma=0.2. Since ground-truth labels were accessible during model selection, Tables 1 and 2 do not measure purely unsupervised clustering performance. The margin over the next-best baseline is small (average rank 3.1579 vs 3.4210 for STCN in Table 1), so selection leakage could plausibly change the ranking. A label-blind selection rule or fixed hyperparameters justified without labels is required.
  2. [Quantitative Analysis; Tables 1 and 2] The baseline comparison is not performed under a common protocol. The paper states that the results in Tables 1 and 2 are sourced from the original papers, except for R-Clustering, so differences in train/test usage, preprocessing, number of clusters, and evaluation scripts are uncontrolled. Without rerunning all baselines in the same framework as VMGAE, or carefully documenting that each published number uses exactly the same protocol, the average ranks and average NMI/RI values in Tables 1 and 2 are not a valid head-to-head comparison.
  3. [Experimental Setup; Tables 1 and 2] No variance or statistical significance is reported: each NMI and RI value appears to come from a single run. Given the stochasticity of the training procedure and the small average-rank differences, the claimed improvements may be within run-to-run noise. Multiple random seeds, standard deviations, and paired significance tests are needed to support the claim of significant improvement.
  4. [Algorithm 1; Eqs. (22)-(24); Appendix F.2] The contribution of the proposed mixture regularizer is not isolated. The training objective explicitly encourages a Gaussian-mixture structure in the latent space through Lreg, and the final clustering step refits a GMM on the same latent embeddings, so part of the reported separability is enforced by construction rather than discovered. The ablation in Appendix F.2 compares different convolutional layers but does not compare VMGAE against GAE or VGAE under identical graph construction and evaluation, nor does it ablate the regularizer by setting lambda=0. Such an ablation is necessary to establish that the mixture regularizer, rather than the graph construction plus GMM post-processing, drives the reported gains.
minor comments (5)
  1. [Graph Construction; Eq. (1)] The text refers to a 'wrapping path' where 'warping path' is intended, and the notation 'dinner' for the elementwise distance should be defined or replaced with a clearer symbol.
  2. [Tables 1 and 2; Appendix F.1] Dataset names are inconsistent: Tables 1 and 2 use 'SonyAIBORobotSurface' while Appendix F.1 uses 'SonyAIBORobotSurface1,' and the header 'u-shaplet' should be 'u-shapelet.' Please standardize these names.
  3. [Introduction] The claim that this is the first work employing a graph autoencoder for time series clustering is too strong without a more exhaustive literature search, especially given the existing graph-based time series clustering work cited later in the paper; the claim should be qualified.
  4. [Application in Finance] The financial application is only qualitative: there are no ground-truth labels, no quantitative clustering evaluation, and no comparison baselines. It should be framed as an illustrative case study rather than a validation of clustering quality.
  5. [Appendix A; Eq. (19)] There are several typesetting issues in Eq. (19): the sum over h is not clearly scoped, and the notation for sigma_h^2 relative to the cluster-specific variance is easy to misread. Please check consistency with Eq. (28) and improve the typesetting.

Circularity Check

1 steps flagged · score 4.0 of 10

The VMGAE derivation itself is not circular; the SOTA claim is partially compromised because hyperparameters are tuned on t-SNE plots colored by ground-truth labels and then evaluated with label-agreement metrics.

  1. fitted input called prediction [Section 4: Experimental Setup; Appendix E, Figure 4 caption]
    "A significant advantage of our method is that we can leverage the latent distribution to tune hyperparameters (as illustrated in Figure 4 in Appendix E). The hyperparameters γ, λ, W, and α were tuned by visualizing the latent distribution of the training set for each dataset separately. ... The visualizations with t-SNE on the dataset DiatomSizeReduction. The colors of the points indicate the actual labels."

    The reported headline results are NMI and RI, which measure agreement between the final GMM clusters and the ground-truth class labels. The same ground-truth labels are used to color the t-SNE plots that are explicitly said to be the basis for per-dataset selection of γ, λ, W, and α. Thus the hyperparameters are fitted, by visual inspection, to a projection that encodes the target labels, and the 'prediction' reported on those datasets is the agreement with those same labels. The comparison is therefore not a label-blind unsupervised benchmark; the claimed best average rank is partly a selection outcome rather than an independent prediction.

full rationale

The formal derivation in the paper is internally coherent and not circular: the ELBO factorization, the q(ci|X,A) update, the MoG regularizer, and the final GMM refit form a standard variational deep-clustering objective. The regularizer is label-free and the benchmark numbers against ground truth are external evidence, so the fact that Lreg encourages Gaussian-mixture-shaped embeddings and the final step fits a GMM is a design choice, not a derivation that reduces to its own conclusion. There are no load-bearing self-citations or imported uniqueness theorems. The central methodological weakness is the model-selection protocol: Section 4 says hyperparameters are tuned by visualizing the latent distribution, and the figure used for this purpose (Figure 4) colors points by the actual labels. Because NMI/RI measure agreement with those same labels, the reported SOTA comparison is contaminated by label-informed selection, which is a fitted-input-called-prediction problem. Appendix F.1 partially mitigates the stronger form of the concern (the authors did not pick the grid point with the best NMI), but it does not remove the label leakage. Accordingly, the derivation receives a modest partial-circularity score; the method still has independent content, including the external Cora experiment and graph-reconstruction objective.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The ledger shows the design choices that carry the method: the WDTW graph with density-based threshold, the mean-field and zero-KL assumptions that reduce the objective to VaDE, the inner-product decoder, and the GMM refit with known K. None of these are invented entities; they are standard modeling assumptions. The largest burden is the per-dataset tuning of γ, λ, W, α by latent visualization, which is not counted as a free parameter in the loss but is a free choice in evaluation.

free parameters (6)
  • WDTW weight slope γ = 0.2 used in main tables; 0.4, 0.7, 1.0 explored in F.1
    Controls how strongly far-apart time indices are penalized in WDTW distance; chosen per dataset by visual inspection of latent space, not derived.
  • WDTW window size W = Not fully reported
    Limits elastic shift between indices; exact values per dataset are not listed, making replication difficult.
  • Graph edge density α = 0.025 to 0.1 explored; exact per-dataset values not tabulated
    Determines threshold δ for adjacency matrix; authors argue α is stable across datasets, but it is a tuned hyperparameter.
  • Regularizer weight λ = 0.1, 0.01, 0.001, 0.0001 explored; exact per-dataset values not tabulated
    Balances graph reconstruction against GMM regularization; set per dataset by visualization.
  • Embedding dimension h = Not specified in main text
    Dimension of latent Z; must be chosen but no default or per-dataset value is given.
  • Number of clusters K = From UCR labels; 5 for stock data via elbow
    For benchmarks K equals true class count; for finance K chosen by elbow method, an additional heuristic.
assumptions (6)
  • domain assumption WDTW distance with weight w[n]=γn and window W defines a meaningful pairwise similarity for time series clustering.
    Invoked at graph construction, Eq. (1); if DTW distances do not capture cluster structure, the graph and all downstream embeddings are misled.
  • ad hoc to paper The edge density α is more stable across datasets than an absolute distance threshold δ.
    Section 'Graph Construction'; this is the paper's stated motivation for its 'novel transformation', and it is a design assumption not derived from data.
  • standard math Mean-field factorization q(Z,c|X,A)=∏ q(z_i|X,A) q(c_i|X,A) is valid.
    Used in Eq. (14); standard variational approximation, but ignores posterior correlations between cluster assignments and embeddings.
  • ad hoc to paper The KL term D_KL(q(c_i|X,A)||p(c_i|z_i)) is set to zero, giving q(c_i|X,A)=p(c_i|z_i).
    Appendix B, Eq. (21); this is an equality assumption that makes the categorical posterior computable, inherited from VaDE and not generally true.
  • domain assumption The decoder p(A|Z)=∏σ(z_i^T z_j) provides an adequate graph reconstruction target.
    Eq. (10); inner-product decoder assumes link probability depends only on embedding dot product, a standard but strong modeling choice.
  • domain assumption A GMM refit on the final latent embeddings produces valid cluster assignments with K equal to the true number of classes.
    Algorithm 1 step 12 and experimental setup; K is taken from the UCR labels, which is standard for benchmark evaluation but requires K known a priori.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Clustering Time Series Data with Gaussian Mixture Embeddings in a Graph Autoencoder Framework." pith.science (2026). https://pith.science/paper/MKCUXNLD

@misc{pith2026241116972,
  author       = {Pith},
  title        = {Pith review of: Clustering Time Series Data with Gaussian Mixture Embeddings in a Graph Autoencoder Framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MKCUXNLD}},
  note         = {Machine review of arXiv:2411.16972}
}
read the original abstract

Time series data analysis is prevalent across various domains, including finance, healthcare, and environmental monitoring. Traditional time series clustering methods often struggle to capture the complex temporal dependencies inherent in such data. In this paper, we propose the Variational Mixture Graph Autoencoder (VMGAE), a graph-based approach for time series clustering that leverages the structural advantages of graphs to capture enriched data relationships and produces Gaussian mixture embeddings for improved separability. Comparisons with baseline methods are included with experimental results, demonstrating that our method significantly outperforms state-of-the-art time-series clustering techniques. We further validate our method on real-world financial data, highlighting its practical applications in finance. By uncovering community structures in stock markets, our method provides deeper insights into stock relationships, benefiting market prediction, portfolio optimization, and risk management.

Figures

Figures reproduced from arXiv: 2411.16972 by the authors.

Figure 1
Figure 1. The general architecture of the Variational Mixture Graph Autoencoder (VMGAE). The dataset [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Graph visualizations of the Symbols dataset, illus [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. (a) Clustering results of the normalized closing [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: The visualizations with t-SNE on the dataset [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

58 extracted references · 39 canonical work pages

  1. [1]

    S.; and Wah, T

    Aghabozorgi, S.; Shirkhorshidi, A. S.; and Wah, T. Y. 2015. Time-series clustering--a decade review. Information systems, 53: 16--38

    work page 2015

  2. [2]

    S.; Geethanjali, N.; and Satyanarayana, B

    Babu, M. S.; Geethanjali, N.; and Satyanarayana, B. 2012. Clustering approach to stock market prediction. International Journal of Advanced Networking and Applications, 3(4): 1281

    work page 2012

  3. [3]

    Cao, D.; Wang, Y.; Duan, J.; Zhang, C.; Zhu, X.; Huang, C.; Tong, Y.; Xu, B.; Bai, J.; Tong, J.; et al. 2020. Spectral temporal graph neural network for multivariate time-series forecasting. Advances in neural information processing systems, 33: 17766--17778

    work page 2020

  4. [4]

    Using Clustering Method to Understand Indian Stock Market Volatility

    Chaudhuri, T. D.; and Ghosh, I. 2016. Using clustering method to understand Indian stock market volatility. arXiv preprint arXiv:1604.05015

    work page Pith review arXiv 2016

  5. [5]

    Close, L.; and Kashef, R. 2020. Combining artificial immune system and clustering analysis: A stock market anomaly detection model. Journal of Intelligent Learning Systems and Applications, 12(04): 83--108

    work page 2020

  6. [6]

    Defferrard, M.; Bresson, X.; and Vandergheynst, P. 2017. Convolutional Neural Networks on Graphs with Fast Localized Spectral Filtering. arXiv:1606.09375

    arXiv 2017

  7. [7]

    Deng, A.; and Hooi, B. 2021. Graph neural network-based anomaly detection in multivariate time series. In Proceedings of the AAAI conference on artificial intelligence, volume 35, 4027--4035

    2021

  8. [8]

    Ding, M.; Chen, Y.; and Bressler, S. L. 2006. Granger causality: basic theory and application to neuroscience. Handbook of time series analysis: recent theoretical developments and applications, 437--460

    work page 2006

Show all 58 references
  1. [9]

    Du, J.; Zhang, S.; Wu, G.; Moura, J. M. F.; and Kar, S. 2018. Topology Adaptive Graph Convolutional Networks. arXiv:1710.10370

    2018 arXiv

  2. [10]

    Fang, Y.; Xu, H.; and Jiang, J. 2020. A survey of time series data visualization research. In IOP Conference Series: Materials Science and Engineering, volume 782, 022013. IOP Publishing

    2020

  3. [11]

    Fey, M.; and Lenssen, J. E. 2019. Fast graph representation learning with PyTorch Geometric. In ICLR Workshop on Representation Learning on Graphs and Manifolds

    2019

  4. [12]

    Guo, X.; Liu, X.; Zhu, E.; and Yin, J. 2017. Deep clustering with convolutional autoencoders. In Neural Information Processing: 24th International Conference, ICONIP 2017, Guangzhou, China, November 14-18, 2017, Proceedings, Part II 24, 373--382. Springer

    2017

  5. [13]

    L.; Ying, R.; and Leskovec, J

    Hamilton, W. L.; Ying, R.; and Leskovec, J. 2018. Inductive Representation Learning on Large Graphs. arXiv:1706.02216

    2018 arXiv

  6. [14]

    Han, S.; and Woo, S. S. 2022. Learning sparse latent graph representations for anomaly detection in multivariate time series. In Proceedings of the 28th ACM SIGKDD Conference on knowledge discovery and data mining, 2977--2986

    2022

  7. [15]

    N.; and McDermid, G

    Hird, J. N.; and McDermid, G. J. 2009. Noise reduction of NDVI time series: An empirical comparison of selected techniques. Remote Sensing of Environment, 113(1): 248--258

    2009

  8. [16]

    Y.; Jiang, N.; and Wang, S

    Huang, X.; Ye, Y.; Xiong, L.; Lau, R. Y.; Jiang, N.; and Wang, S. 2016. Time series k-means: A new k-means type smooth subspace clustering for time series data. Information Sciences, 367-368: 1--13

    2016

  9. [17]

    Ismail Fawaz, H.; Forestier, G.; Weber, J.; Idoumghar, L.; and Muller, P.-A. 2019. Deep learning for time series classification: a review. Data mining and knowledge discovery, 33(4): 917--963

    2019

  10. [18]

    Jiang, Z.; Zheng, Y.; Tan, H.; Tang, B.; and Zhou, H. 2016. Variational deep embedding: An unsupervised and generative approach to clustering. arXiv preprint arXiv:1611.05148

    2016 arXiv

  11. [19]

    Jorge, M.-B.; and Rub \'e n, C. 2024. Time series clustering with random convolutional kernels. Data Mining and Knowledge Discovery, 1--27

    2024

  12. [20]

    N.; and Welling, M

    Kipf, T. N.; and Welling, M. 2016 a . Semi-supervised classification with graph convolutional networks. arXiv preprint arXiv:1609.02907

    2016 arXiv

  13. [21]

    N.; and Welling, M

    Kipf, T. N.; and Welling, M. 2016 b . Variational graph auto-encoders. arXiv preprint arXiv:1611.07308

    2016 arXiv

  14. [22]

    Lee, J.; Kim, D.; and Sim, S. 2024. Temporal Multi-features Representation Learning-Based Clustering for Time-Series Data. IEEE Access

    2024

  15. [23]

    S.; Mah, D

    Li, G.; Choi, B.; Xu, J.; Bhowmick, S. S.; Mah, D. N.-y.; and Wong, G. L.-H. 2022. Autoshape: An autoencoder-shapelet approach for time series clustering. arXiv preprint arXiv:2208.04313

    2022 arXiv

  16. [24]

    W.; Wu, K.; and Wan, Y

    Li, H.; Liu, J.; Yang, Z.; Liu, R. W.; Wu, K.; and Wan, Y. 2020. Adaptively constrained dynamic time warping for time series classification and clustering. Information Sciences, 534: 97--116

    2020

  17. [25]

    F.; and Hamdi, S

    Li, P.; Boubrahimi, S. F.; and Hamdi, S. M. 2021. Graph-based clustering for time series data. In 2021 IEEE International Conference on Big Data (Big Data), 4464--4467. IEEE

    2021

  18. [26]

    Liao, T. W. 2005. Clustering of time series data—a survey. Pattern recognition, 38(11): 1857--1874

    2005

  19. [27]

    Lin, J.; Williamson, S.; Borne, K.; and DeBarr, D. 2012. Pattern recognition in time series. Advances in machine learning and data mining for astronomy, 1(617-645): 3

    2012

  20. [28]

    Ma, Q.; Li, S.; Zhuang, W.; Wang, J.; and Zeng, D. 2020. Self-supervised time series clustering with model-based dynamics. IEEE Transactions on Neural Networks and Learning Systems, 32(9): 3942--3955

    2020

  21. [29]

    Ma, Q.; Zheng, J.; Li, S.; and Cottrell, G. W. 2019. Learning representations for time series clustering. Advances in neural information processing systems, 32

    2019

  22. [30]

    Mudelsee, M. 2019. Trend analysis of climate time series: A review of methods. Earth-science reviews, 190: 310--322

    2019

  23. [31]

    Olive, X.; Basora, L.; Viry, B.; and Alligier, R. 2020. Deep trajectory clustering with autoencoders. In ICRAT 2020, 9th International Conference for Research in Air Transportation

    2020

  24. [32]

    Pan, S.; Hu, R.; Long, G.; Jiang, J.; Yao, L.; and Zhang, C. 2019. Adversarially Regularized Graph Autoencoder for Graph Embedding. arXiv:1802.04407

    2019 arXiv

  25. [33]

    Paparrizos, J.; and Gravano, L. 2015. k-shape: Efficient and accurate clustering of time series. In Proceedings of the 2015 ACM SIGMOD international conference on management of data, 1855--1870

    2015

  26. [34]

    Paszke, A.; Gross, S.; Massa, F.; Lerer, A.; Bradbury, J.; Chanan, G.; Killeen, T.; Lin, Z.; Gimelshein, N.; Antiga, L.; et al. 2019. PyTorch: An imperative style, high-performance deep learning library. Advances in Neural Information Processing Systems, 32: 8024--8035

    2019

  27. [35]

    Sakoe, H. 1978. Dynamic programming algorithm optimization for spoken word recognition. IEEE Transactions on Acoustics, Speech, and Signal Processing, 26: 159--165

    1978

  28. [36]

    Shah, D.; Isah, H.; and Zulkernine, F. 2019. Stock market analysis: A review and taxonomy of prediction techniques. International Journal of Financial Studies, 7(2): 26

    2019

  29. [37]

    M.; Luo, S.; Shabbir, S.; Hameed, I

    Shaukat, K.; Alam, T. M.; Luo, S.; Shabbir, S.; Hameed, I. A.; Li, J.; Abbas, S. K.; and Javed, U. 2021. A review of time-series anomaly detection techniques: A step to future perspectives. In Advances in Information and Communication: Proceedings of the 2021 Future of Informa...

    2021

  30. [38]

    Siuly, S.; Li, Y.; and Zhang, Y. 2016. EEG signal analysis and classification. IEEE Trans Neural Syst Rehabilit Eng, 11: 141--144

    2016

  31. [39]

    Song, C.; Lin, Y.; Guo, S.; and Wan, H. 2020. Spatial-temporal synchronous graph convolutional networks: A new framework for spatial-temporal network data forecasting. In Proceedings of the AAAI conference on artificial intelligence, volume 34, 914--921

    2020

  32. [40]

    Y.; Saha, H.; Jacoby, M.; Henze, G.; and Sarkar, S

    Tan, S. Y.; Saha, H.; Jacoby, M.; Henze, G.; and Sarkar, S. 2020. Granger causality based hierarchical time series clustering for state estimation. IFAC-PapersOnLine, 53(2): 524--529

    2020

  33. [41]

    Thorndike, R. L. 1953. Who belongs in the family? Psychometrika, 18(4): 267--276

    1953

  34. [42]

    F.; Hadjout, D.; Sebaa, A.; Mart \' nez- \'A lvarez, F.; and Troncoso, A

    Torres, J. F.; Hadjout, D.; Sebaa, A.; Mart \' nez- \'A lvarez, F.; and Troncoso, A. 2021. Deep learning for time series forecasting: a survey. Big Data, 9(1): 3--21

    2021

  35. [43]

    Ulanova, L.; Begum, N.; and Keogh, E. 2015. Scalable clustering of time series with u-shapelets. In Proceedings of the 2015 SIAM international conference on data mining, 900--908. SIAM

    2015

  36. [44]

    van der Maaten, L.; and Hinton, G. 2008. Visualizing Data using t-SNE. Journal of Machine Learning Research, 9(86): 2579--2605

    2008

  37. [45]

    Veličković, P.; Cucurull, G.; Casanova, A.; Romero, A.; Liò, P.; and Bengio, Y. 2018. Graph Attention Networks. arXiv:1710.10903

    2018 arXiv

  38. [46]

    Wu, F.; Souza, A.; Zhang, T.; Fifty, C.; Yu, T.; and Weinberger, K. Q. 2019. Simplifying Graph Convolutional Networks. In Proceedings of the 36th International Conference on Machine Learning, 6861--6871

    2019

  39. [47]

    Xi, W.; Jain, A.; Zhang, L.; and Lin, J. 2023. Lb-simtsc: An efficient similarity-aware graph neural network for semi-supervised time series classification. arXiv preprint arXiv:2301.04838

    2023 arXiv

  40. [48]

    Xie, J.; Girshick, R.; and Farhadi, A. 2016. Unsupervised deep embedding for clustering analysis. In International conference on machine learning, 478--487. PMLR

    2016

  41. [49]

    Yang, J.; and Leskovec, J. 2011. Patterns of temporal variation in online media. In Proceedings of the fourth ACM international conference on Web search and data mining, 177--186

    2011

  42. [50]

    Yu, B.; Yin, H.; and Zhu, Z. 2017. Spatio-temporal graph convolutional networks: A deep learning framework for traffic forecasting. arXiv preprint arXiv:1709.04875

    2017 arXiv

  43. [51]

    Zha, D.; Lai, K.-H.; Zhou, K.; and Hu, X. 2022. Towards similarity-aware time-series classification. In Proceedings of the 2022 SIAM International Conference on Data Mining (SDM), 199--207. SIAM

    2022

  44. [52]

    Zhang, Q.; Wu, J.; Yang, H.; Tian, Y.; and Zhang, C. 2016. Unsupervised feature learning from time series. In IJCAI, 2322--2328. New York, USA

    2016

  45. [53]

    Zhang, Q.; Wu, J.; Zhang, P.; Long, G.; and Zhang, C. 2018. Salient subsequence learning for time series clustering. IEEE transactions on pattern analysis and machine intelligence, 41(9): 2193--2207

    2018

  46. [54]

    Zhao, H.; Wang, Y.; Duan, J.; Huang, C.; Cao, D.; Tong, Y.; Xu, B.; Bai, J.; Tong, J.; and Zhang, Q. 2020. Multivariate time-series anomaly detection via graph attention network. In 2020 IEEE international conference on data mining (ICDM), 841--850. IEEE

    2020

  47. [55]

    Zhao, J.; and Itti, L. 2018. shapeDTW: Shape dynamic time warping. Pattern Recognition, 74: 171--184

    2018

  48. [56]

    Zhao, J.; Zhu, N.; and Lu, S. 2009. Productivity model in hot and humid environment based on heat tolerance time analysis. Building and environment, 44(11): 2202--2207

    2009

  49. [57]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all...

  50. [58]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

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