Pith. sign in

REVIEW 3 major objections 5 minor 39 references

CASC: Causal Adversarial Subspace Clustering for Multivariate Spatiotemporal Data

T0 review · 3 major / 5 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read This paper claims that adding a causal consistency term to deep subspace clustering makes clusters reflect latent causal processes, not just feature similarity, transforming the method into a causal-temporal regime discovery framework.

desk verdict A real piece of architecture and two new loss terms, but the causal-regime claim is a self-consistency regularizer dressed up as causal discovery, and the experiments don't rescue it. read the letter →

arxiv 2607.21088 v1 pith:VEFV3PCF submitted 2026-07-23 cs.LG

Francis Ndikum Nji , Vandana Janeja , Jianwu Wang This is my paper
classification cs.LG MSC 62H3068T0562M10
keywords deepsubspaceclusteringspatiotemporaldatacausaldiscoveryadversariallearninggraphattentiontransformertemporalregimeself-expressivenetworkenergy-baseddiscriminator
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

CASC is a deep subspace clustering framework for multivariate spatiotemporal data that joins adversarial representation learning with two new objectives. The first objective, a Causal Subspace Preservation Loss, aligns the learned self-expression coefficients with a latent causal adjacency matrix, pushing clusters to group samples by shared driving mechanisms rather than mere geometric similarity. The second, a Dynamic Temporal Subspace Evolution Loss, allows subspaces to change gradually over time so the model can capture regime transitions in nonstationary systems. The paper argues that these components, together with a subspace-aware energy-based discriminator, turn deep subspace clustering from a correlation-driven tool into a causal-temporal regime discovery method, and reports improved internal cluster quality on three climate reanalysis datasets.

What carries the argument

The load-bearing mechanism is the Causal Subspace Preservation Loss, which augments the standard self-expression objective with a Frobenius-norm alignment term ||A_causal - CC^T||_F^2. Here C is the self-expression coefficient matrix, CC^T induces an affinity graph, and A_causal is a causal adjacency matrix estimated from latent features using a causal discovery mechanism such as Neural Granger Causality or PCMCI+. This alignment is what converts the clustering objective from purely geometric self-expressiveness into one that rewards clusters that respect latent causal structure. The Dynamic Temporal Subspace Evolution Loss extends the machinery by replacing the static C with a sequence {C_t

What would settle it

On synthetic multivariate spatiotemporal data with known causal regimes, if CASC's clusters fail to align with the true regime boundaries (e.g., near-random adjusted Rand index), the causal claim is disproven. A simpler check: replace A_causal with a random matrix of the same density on a real dataset; if clustering quality (silhouette, DB) does not degrade significantly, the causal loss is not the source of the reported improvement.

Watch

Extended reading notes

Core claim

The central claim is that the clusters produced by deep subspace clustering become causally meaningful and temporally coherent when the self-expression matrix C is explicitly aligned with an estimated causal adjacency matrix A_causal, and when consecutive self-expression matrices are regularized to change slowly. The paper proposes the Causal Subspace Preservation Loss, L_CSP = ||Z - ZC||_F^2 + λ_c ||A_causal - CC^T||_F^2 + λ_r||C||_p, which biases the learned subspaces toward those consistent with latent causal relationships, and the Dynamic Temporal Subspace Evolution Loss, L_DTSE = Σ_t ||C_t - C_{t-1}||_1, which permits gradual subspace drift. Along with a Subspace-Aware Energy-based Temp

Load-bearing premise

The causal adjacency matrix A_causal, estimated from the same latent features being trained and never validated against any ground-truth causal graph, is assumed to reflect the true latent causal relationships of the physical system.

Editorial extensions

If this is right

  • If correct, clusters extracted from climate reanalysis data would correspond to distinct atmospheric or oceanic regimes driven by shared physical mechanisms, improving interpretability for downstream scientific analysis.
  • The temporal evolution loss provides a principled way to track how latent subspaces (and hence regimes) shift over time, allowing the model to detect transition boundaries in nonstationary systems.
  • The energy-based discriminator scales as O(K d r) in trainable parameters for K clusters, latent dimension d, and subspace rank r, offering a more parameter-efficient alternative to deep discriminator networks.
  • The causal alignment term can, in principle, be grafted onto any self-expressive subspace clustering method, giving a general recipe to inject causal structure into unsupervised learning.
  • Jointly optimizing causal consistency, temporal smoothness, reconstruction, and adversarial subspace quality yields clusters that are simultaneously compact, well-separated, and temporally coherent, as demonstrated on three real-world datasets.

Reading between the lines

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

  • A natural extension is to run CASC with several distinct causal discovery algorithms (e.g., Granger causality, PCMCI+, transfer entropy) and compare resulting clusters; substantial sensitivity would suggest the causal alignment is not a stable identifier of latent physical processes.
  • The cleanest test of the causal claim would be on synthetic multivariate spatiotemporal data with known ground-truth causal regime boundaries; if CASC's clusters do not match those boundaries, the 'causal regime discovery' claim reduces to a regularized clustering improvement.
  • The DTSE regularizer could be detached and applied to any time-varying clustering method, potentially improving temporal coherence without the causal machinery, which would isolate its contribution.
  • Because A_causal is derived from the same latent features being optimized, the CSP loss may create a self-consistency loop that encourages clustering to match the discovered graph regardless of the graph's external validity; a controlled experiment that corrupts A_causal (e.g., random permutation) would reveal whether clusters depend on causal content or merely on the regularization shape.
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

3 major / 5 minor

Summary. The paper proposes CASC, an end-to-end deep subspace clustering framework for multivariate spatiotemporal data, combining a FAConvLSTM-based U-Net autoencoder, a bidirectional temporal graph attention bottleneck, a DEC-style clustering head, a self-expressive layer, and a subspace-aware energy-based discriminator. Two new losses are introduced: a Causal Subspace Preservation (CSP) loss that aligns the self-expression Gram matrix C C^T with a 'latent causal adjacency' A_causal, and a Dynamic Temporal Subspace Evolution (DTSE) loss that penalizes differences between consecutive self-expression matrices. Experiments on ERA5, CARRA, and NCAR reanalysis data compare CASC with five deep clustering baselines using six internal cluster validity metrics, plus an ablation study on ERA5.

Significance. If substantiated, incorporating causal structure into deep subspace clustering for spatiotemporal data would be a meaningful contribution. The proposed architecture is intricate and the ablation study suggests each module contributes something. The paper also targets a relevant application area. However, the central causal claim is not validated: the causal adjacency matrix is never concretely defined, estimated, or compared against any ground truth, and it is derived from the very latent features being optimized, making the CSP loss a self-consistency regularizer rather than an independent causal constraint. The experimental section relies entirely on internal clustering metrics with no repeated runs or significance tests, and the claimed 'consistently superior' performance is not supported by Table I, where baseline methods beat CASC on several metrics. The significance of the contributions therefore remains unestablished as presented.

major comments (3)
  1. [Section VI, Table I and Abstract] The causal core of the method is not well-posed. A_causal is described only as 'estimated from latent features using a causal discovery mechanism such as Neural Granger Causality, PCMCI+, Transfer Entropy, or Convergent Cross Mapping,' but no algorithm is selected, implemented, or cited, and A_causal is never reported or compared with any external causal ground truth. Since the latent features Z are exactly the variables being optimized by the network, A_causal is a moving target: the loss can be reduced either by improving the causal estimate or by distorting Z to make the output of any causal discovery procedure consistent with C C^T. Without an external causal anchor, Eq. (7) is a self-consistency regularizer, not a causal constraint. There is also a dimensional/identifiability issue: causal adjacency in multivariate spatiotemporal data is naturally defined between variables (channels
  2. [Section VI, Table I and Abstract] The abstract claims 'substantial improvements in clustering accuracy, interpretability, and robustness,' but no ground-truth labels are used anywhere; all six metrics are internal cluster validity indices, so 'accuracy' is never actually measured. There are no repeated runs, no standard deviations, and no significance tests, making it impossible to assess whether the reported differences are meaningful. Furthermore, Table I does not support the 'consistently superior' claim: DSC has higher CH on ERA5 and CARRA, DEC has lower RMSE on CARRA, and DTC has lower DB and higher I-CD on NCAR. The paper needs multiple random seeds with mean±std, statistical significance tests, and ideally at least one dataset or experiment with external labels or known regimes to substantiate the accuracy and robustness claims.
  3. [Section V-A.7 / Eq. (6)] The loss in Eq. (6) is undefined as written because the candidate causal discovery methods produce very different outputs: Neural Granger and PCMCI+ yield directed graphs over variables, while Transfer Entropy and Convergent Cross Mapping can be computed at various levels of aggregation. The manuscript does not specify how any of these is applied to the latent tensor Z, how the resulting graph is thresholded or weighted, or how it is symmetrized and normalized to be comparable with C C^T, which is a Gram matrix of self-expression coefficients. This is a reproducibility blocker: a reader cannot implement or evaluate the central new term. The authors should provide the exact construction, including the estimator, the preprocessing, and the hyperparameters used in the reported experiments.
minor comments (5)
  1. [Section V and VI] Figure numbering is inconsistent: Section V refers to 'Figure 3' for the architecture, while Section VI-D uses 'Figure 3' for the clustering visualization. The architecture appears to be Figure 1. Please renumber figures and update references.
  2. [Section V-A.2 and V-A.7] Notation for the latent variable is inconsistent: Z is defined as R^{T×d} in the encoder and then as R^{d×N} in the CSP loss. Also z_t is used both for the t-th timestep and for the i-th sample. Clarify whether N counts time steps or samples, and align the notation throughout.
  3. [Section VI-A] The experimental setup description is confusing: the text mentions AWS cloud, Google Colab, and local macOS hardware. Please specify exactly where each experiment was run and which GPU was used for the reported results.
  4. [Section VI-D] There are repeated typos: 'Calinski-Harabas' should be 'Calinski-Harabasz,' 'applicatins' should be 'applications,' and Table II headers 'CASCCSP', 'CASCDT SE', 'CASCSET D' need spaces for readability.
  5. [References] The causal discovery methods named in Section V-A.7 (Neural Granger Causality, PCMCI+, Transfer Entropy, Convergent Cross Mapping) are not cited. Please add appropriate references so the reader can locate the methods.

Circularity Check

1 steps flagged · score 6.0 of 10

CSP loss aligns C C^T with A_causal estimated from the same latent features Z, making the 'causal regime' claim a self-consistency constraint rather than an independent causal discovery result.

  1. self definitional [Section V-A.7, Eqs. (6)-(7)]
    "Let A_causal ∈ R^{N×N} denote a causal adjacency matrix estimated from latent features using a causal discovery mechanism such as Neural Granger Causality, PCMCI+, Transfer Entropy, or Convergent Cross Mapping. The affinity structure induced by the self-expression coefficients is represented by S=CC^T. The proposed causal consistency objective is defined as: L_causal = ||A_causal − CC^T||_F^2. (6)"

    Both A_causal and C are functions of the same latent representation Z that the network is training. A_causal is 'estimated from latent features' Z, while C is the self-expression coefficient matrix obtained from the same Z via the self-expressiveness objective. Minimizing (6) therefore only forces two projections of the same learned Z to agree; it injects no independent causal information. The paper then interprets this internal consistency as making clusters 'causally meaningful' and 'reflect underlying causal processes,' but no external causal graph or downstream causal validation is provided. The causal signal is defined by the very representation it is supposed to supervise, so the claimed causal-regime discovery reduces to a self-consistency regularizer.

full rationale

The central causal claim of CASC rests on the Causal Subspace Preservation loss, Eq. (6)-(7). The paper says A_causal is 'estimated from latent features' using a causal discovery mechanism, and the self-expression coefficient matrix C is learned from the same latent features via L_self = ||Z - ZC||_F^2. Aligning C C^T with A_causal is thus a consistency term between two quantities computed from the same learned representation, not an independent causal oracle. No specific causal discovery algorithm is implemented, no A_causal is reported, and no external causal ground truth is used to test whether clusters correspond to physical causal regimes. Consequently, the abstract's claim that CASC 'discovers latent regimes... that correspond to underlying causal processes' is not supported by an independent causal signal; it is a self-supervised alignment objective. The rest of the framework—adversarial training, DTSE temporal smoothness, reconstruction, and internal clustering metrics—contains legitimate non-circular components and empirical comparisons, but those do not validate the causal semantics. The self-citations to prior FAConvLSTM and DSC work are architectural/baseline references, not load-bearing circular justifications. Overall, one core 'causal' component reduces by construction to self-consistency, meriting a 6 rather than a higher score.

Assumptions & free parameters 7 free parameters · 5 assumptions · 1 invented entities

The central claim depends on many unspecified hyperparameters (loss weights, margin, rank, schedules, architecture dims) and on an unvalidated causal adjacency matrix estimated from the learned representation itself. The subspace clustering and self-expression axioms are standard; the causal-discovery premise and the internal-metric validity premise are the most fragile, since they carry the paper's headline claims without independent support.

free parameters (7)
  • Loss weights λ1-λ6 (Eq. 9)
    Control the balance between reconstruction, adversarial, clustering, self-expression, causal, and temporal-evolution losses; values are not given in the paper.
  • Discriminator margin m
    Hinge-loss margin in the energy-based discriminator; unspecified.
  • Subspace dimensionality r
    Rank of each learned subspace basis U_k; unspecified.
  • Number of clusters K per dataset = ERA5: 7, CARRA: 5, NCAR: 7
    Chosen by elbow method on the same data that is evaluated, so cluster count is fit to the target data.
  • Temperature schedule τ and Student-t dof α
    DEC-style sharpening schedule and degrees-of-freedom; unspecified.
  • Patch sizes (h_p, w_p), attention heads, latent dim D
    Architectural choices for patchification and Bi-TGAT; not specified.
  • Causal discovery algorithm choice
    Paper lists Neural Granger Causality, PCMCI+, Transfer Entropy, or Convergent Cross Mapping as options but never selects one, so the causal signal source is a free choice.
assumptions (5)
  • domain assumption Data lies in a union of M low-dimensional subspaces of a Hilbert/Banach space (Section IV).
    Standard subspace clustering assumption; not directly verified for climate data.
  • standard math Self-expressiveness: each latent point is a linear combination of others in the same subspace (Section V-A.7).
    Core premise of all self-expressive subspace clustering methods.
  • ad hoc to paper A_causal, estimated from latent features by an unspecified causal discovery method, captures true latent causal relationships (Section V-A.7).
    Unvalidated and load-bearing for the 'causal' claim.
  • domain assumption Internal cluster validation metrics (Silhouette, DB, CH, etc.) are meaningful proxies for clustering quality on unlabeled climate data (Section VI-C).
    No ground truth or external validation is used.
  • domain assumption Time steps can be partitioned into K coherent regimes, with K chosen by elbow method.
    Assumes discrete regime structure for continuous climate dynamics.
invented entities (1)
  • Latent causal adjacency matrix A_causal over time steps
    purpose: Acts as the supervision target for the Causal Subspace Preservation Loss, claiming to encode causal relationships among latent features.
    It is 'estimated' by an unspecified causal discovery method from the same latent features being trained; no external causal graph or falsifiable downstream prediction is provided, making it a postulated latent construct with no independent handle.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CASC: Causal Adversarial Subspace Clustering for Multivariate Spatiotemporal Data." pith.science (2026). https://pith.science/paper/VEFV3PCF

@misc{pith2026260721088,
  author       = {Pith},
  title        = {Pith review of: CASC: Causal Adversarial Subspace Clustering for Multivariate Spatiotemporal Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VEFV3PCF}},
  note         = {Machine review of arXiv:2607.21088}
}
read the original abstract

Deep subspace clustering plays a critical role in applications involving multivariate spatiotemporal data, such as sea ice monitoring, disease spread analysis, and tracking neuro-degeneration over time. Despite recent advances, existing methods primarily rely on geometric self-expressiveness, assume static subspace structures, and often fail to capture causal dependencies, local spatial interactions, and long-range temporal dynamics inherent in complex spatiotemporal systems. To address these limitations, we propose a novel Causal Adversarial Subspace Clustering (CASC) framework for discovering evolving latent regimes in high-dimensional spatiotemporal data. CASC integrates a U-Net-inspired deep adversarial clustering architecture with stacked FAConvLSTM layers to preserve spatial and temporal structure while learning robust latent representations. A graph attention transformer-based self-expressive network is introduced to jointly model local spatial relationships, global dependencies, and long-range temporal interactions. Furthermore, we propose two new learning objectives: (1) a Causal Subspace Preservation Loss that aligns self-expression coefficients with latent causal relationships, encouraging clusters to reflect underlying causal processes rather than simple feature similarity, and (2) a Dynamic Temporal Subspace Evolution Loss that captures evolving subspace structures and temporal regime transitions in nonstationary environments. Together, these components transform deep subspace clustering from a correlation-driven paradigm into a causal-temporal regime discovery framework.

Figures

Figures reproduced from arXiv: 2607.21088 by the authors.

Figure 1
Figure 1. Architecture of Causal Adversarial Subspace Clustering for Multivariate Spatiotemporal Data (CASC) framework. CASC is composed of a deep [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. 3D Spatiotemporal View of Surface Pressure (SP) [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. CASC Clustering Visualization [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

39 extracted references · 4 linked inside Pith

  1. [1]

    Investigating the spatio-temporal distribution of gravity wave potential energy over the equatorial region using the era5 reanalysis data,

    S.-S. Yang, C.-J. Pan, and U. Das, “Investigating the spatio-temporal distribution of gravity wave potential energy over the equatorial region using the era5 reanalysis data,”Atmosphere, vol. 12, no. 3, p. 311, 2021

    2021

  2. [2]

    The intrinsic dimension of images and its impact on learning,

    P. Pope, C. Zhu, A. Abdelkader, M. Goldblum, and T. Goldstein, “The intrinsic dimension of images and its impact on learning,”arXiv preprint arXiv:2104.08894, 2021

    arXiv 2021

  3. [3]

    Deep convolutional recurrent autoen- coders for learning low-dimensional feature dynamics of fluid systems,

    F. J. Gonzalez and M. Balajewicz, “Deep convolutional recurrent autoen- coders for learning low-dimensional feature dynamics of fluid systems,” arXiv preprint arXiv:1808.01346, 2018

    arXiv 2018

  4. [4]

    Ldmnet: Low dimensional manifold regularized neural networks,

    W. Zhu, Q. Qiu, J. Huang, R. Calderbank, G. Sapiro, and I. Daubechies, “Ldmnet: Low dimensional manifold regularized neural networks,” in Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2743–2751, 2018

    2018

  5. [5]

    Principal component analysis (pca),

    T. Kurita, “Principal component analysis (pca),”Computer Vision: A Reference Guide, pp. 1–4, 2019

    2019

  6. [6]

    Stochastic sparse subspace clustering,

    Y . Chen, C.-G. Li, and C. You, “Stochastic sparse subspace clustering,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 4155–4164, 2020

    2020

  7. [7]

    Robust recovery of subspace structures by low-rank representation,

    G. Liu, Z. Lin, S. Yan, J. Sun, Y . Yu, and Y . Ma, “Robust recovery of subspace structures by low-rank representation,”IEEE transactions on pattern analysis and machine intelligence, vol. 35, no. 1, pp. 171–184, 2012

    2012

  8. [8]

    A selective overview of recent ad- vances in spectral clustering and their applications,

    Y . Xu, A. Srinivasan, and L. Xue, “A selective overview of recent ad- vances in spectral clustering and their applications,”Modern Statistical Methods for Health Research, pp. 247–277, 2021

    2021

Show all 39 references
  1. [9]

    Survey of spectral clustering based on graph theory,

    L. Ding, C. Li, D. Jin, and S. Ding, “Survey of spectral clustering based on graph theory,”Pattern Recognition, vol. 151, p. 110366, 2024

    2024

  2. [10]

    Deep spectral cluster- ing using dual autoencoder network,

    X. Yang, C. Deng, F. Zheng, J. Yan, and W. Liu, “Deep spectral cluster- ing using dual autoencoder network,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 4066–4075, 2019

    2019

  3. [11]

    Multi-scale fusion subspace clustering using similarity constraint,

    Z. Dang, C. Deng, X. Yang, and H. Huang, “Multi-scale fusion subspace clustering using similarity constraint,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp. 6658–6667, 2020

    2020

  4. [12]

    Lrsc: Learning representations for subspace clustering,

    C. Li, C. Yang, B. Liu, Y . Yuan, and G. Wang, “Lrsc: Learning representations for subspace clustering,” inProceedings of the AAAI conference on Artificial Intelligence, vol. 35, pp. 8340–8348, 2021

    2021

  5. [13]

    Deep subspace clus- tering networks,

    P. Ji, T. Zhang, H. Li, M. Salzmann, and I. Reid, “Deep subspace clus- tering networks,”Advances in neural information processing systems, vol. 30, 2017

    2017

  6. [14]

    Scalable deep k- subspace clustering,

    T. Zhang, P. Ji, M. Harandi, R. Hartley, and I. Reid, “Scalable deep k- subspace clustering,” inAsian Conference on Computer Vision, pp. 466– 481, Springer, 2018

    2018

  7. [15]

    Large-scale subspace clustering via k-factorization,

    J. Fan, “Large-scale subspace clustering via k-factorization,” inProceed- ings of the 27th ACM SIGKDD conference on knowledge discovery & data mining, pp. 342–352, 2021

    2021

  8. [16]

    Learning a self-expressive network for subspace clustering,

    S. Zhang, C. You, R. Vidal, and C.-G. Li, “Learning a self-expressive network for subspace clustering,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 12393– 12403, 2021

    2021

  9. [17]

    U-net: Convolutional networks for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional networks for biomedical image segmentation,” inMedical image computing and computer-assisted intervention–MICCAI 2015: 18th international con- ference, Munich, Germany, October 5-9, 2015, proceedings, part III 18, ...

    2015

  10. [18]

    Faconvlstm: Factorized-attention convlstm for efficient feature extraction in multivariate climate data,

    F. N. Nji and J. Wang, “Faconvlstm: Factorized-attention convlstm for efficient feature extraction in multivariate climate data,”arXiv preprint arXiv:2601.10914, 2026

    2026

  11. [19]

    Deep self-representative subspace clustering network,

    S. Baek, G. Yoon, J. Song, and S. M. Yoon, “Deep self-representative subspace clustering network,”Pattern Recognition, vol. 118, p. 108041, 2021

    2021

  12. [20]

    Deep double self-expressive subspace clustering,

    L. Zhao, Y . Ma, S. Chen, and J. Zhou, “Deep double self-expressive subspace clustering,” inICASSP 2023-2023 IEEE International Confer- ence on Acoustics, Speech and Signal Processing (ICASSP), pp. 1–5, IEEE, 2023

    2023

  13. [21]

    Deep adversarial subspace clustering,

    P. Zhou, Y . Hou, and J. Feng, “Deep adversarial subspace clustering,” inProceedings of the IEEE conference on computer vision and pattern recognition, pp. 1596–1604, 2018

    2018

  14. [22]

    Clustergan: Latent space clustering in generative adversarial networks,

    S. Mukherjee, H. Asnani, E. Lin, and S. Kannan, “Clustergan: Latent space clustering in generative adversarial networks,” inProceedings of the AAAI conference on artificial intelligence, vol. 33, pp. 4610–4617, 2019

    2019

  15. [23]

    Gan- based enhanced deep subspace clustering networks,

    Z. Yu, Z. Zhang, W. Cao, C. Liu, C. P. Chen, and H.-S. Wong, “Gan- based enhanced deep subspace clustering networks,”IEEE Transactions on Knowledge and Data Engineering, vol. 34, no. 7, pp. 3267–3281, 2020

    2020

  16. [24]

    Unsupervised deep embedding for clustering analysis,

    J. Xie, R. Girshick, and A. Farhadi, “Unsupervised deep embedding for clustering analysis,” inInternational conference on machine learning, pp. 478–487, PMLR, 2016

    2016

  17. [25]

    Deep spatiotemporal clustering: A temporal clustering approach for multi-dimensional climate data,

    O. Faruque, F. N. Nji, M. Cham, R. M. Salvi, X. Zheng, and J. Wang, “Deep spatiotemporal clustering: A temporal clustering approach for multi-dimensional climate data,”arXiv preprint arXiv:2304.14541, 2023

    2023 arXiv

  18. [26]

    Deep learning for precipitation nowcasting: A benchmark and a new model,

    X. Shi, Z. Gao, L. Lausen, H. Wang, D.-Y . Yeung, W.-K. Wong, and W.-c. Woo, “Deep learning for precipitation nowcasting: A benchmark and a new model,”Advances in Neural Information Processing Systems (NeurIPS), vol. 30, 2017

    2017

  19. [27]

    Deep learning for arctic sea ice prediction,

    L. Wang, K. A. Scott, L. Xu, and J. R. Key, “Deep learning for arctic sea ice prediction,”Remote Sensing, vol. 11, no. 24, p. 2903, 2019

    2019

  20. [28]

    Spatiotemporal deep learning for regional climate downscaling,

    Y . Liu, E. Racah, J. Correa, K. Kunkel, W. Li, and M. Wehner, “Spatiotemporal deep learning for regional climate downscaling,”Nature Communications, vol. 11, p. 4728, 2020

    2020

  21. [29]

    Auto-encoding variational bayes,

    D. P. Kingma and M. Welling, “Auto-encoding variational bayes,”arXiv preprint arXiv:1312.6114, 2013

    2013 arXiv

  22. [30]

    climate.copernicus.eu

    Copernicus Climate Change Service (C3S), “CARRA.” https://cds. climate.copernicus.eu

  23. [31]

    Era5 hourly data on single levels from 1940 to present

    ECMWF, Copernicus Climate Change Service, “Era5 hourly data on single levels from 1940 to present.” https://cds.climate.copernicus.eu/ cdsapp##!/dataset/reanalysis-era5-single-levels?tab=overview, 2021

    1940

  24. [32]

    Daily atmospheric observations

    NCEP/NCAR, “Daily atmospheric observations.” https://www.psl.noaa. gov/data

  25. [33]

    Deep tem- poral clustering: Fully unsupervised learning of time-domain features,

    N. Sai Madiraju, S. M. Sadat, D. Fisher, and H. Karimabadi, “Deep tem- poral clustering: Fully unsupervised learning of time-domain features,” arXiv e-prints, pp. arXiv–1802, 2018

    2018

  26. [34]

    Cluster quality analysis using silhouette score,

    K. R. Shahapure and C. Nicholas, “Cluster quality analysis using silhouette score,” in2020 IEEE 7th international conference on data science and advanced analytics (DSAA), pp. 747–748, IEEE, 2020

    2020

  27. [35]

    Pdbi: A partitioning davies-bouldin index for clustering evaluation,

    F. Ros, R. Riad, and S. Guillaume, “Pdbi: A partitioning davies-bouldin index for clustering evaluation,”Neurocomputing, vol. 528, pp. 178–199, 2023

    2023

  28. [36]

    An improved index for clustering validation based on silhouette index and calinski-harabasz index,

    X. Wang and Y . Xu, “An improved index for clustering validation based on silhouette index and calinski-harabasz index,” inIOP Conference Series: Materials Science and Engineering, vol. 569, p. 052024, IOP Publishing, 2019

    2019

  29. [37]

    Cluster analysis, wiley,

    B. Everitt, S. Landau, M. Leese, and D. Stahl, “Cluster analysis, wiley,” Chichester, UK, 2011

    2011

  30. [38]

    D. C. Montgomery and G. C. Runger,Applied statistics and probability for engineers. John wiley & sons, 2010

    2010

  31. [39]

    Advantages of the mean absolute error (mae) over the root mean square error (rmse) in assessing average model performance,

    C. J. Willmott and K. Matsuura, “Advantages of the mean absolute error (mae) over the root mean square error (rmse) in assessing average model performance,”Climate research, vol. 30, no. 1, pp. 79–82, 2005

    2005

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.