Pith. sign in

REVIEW 3 major objections 4 minor 32 references

Deep Structured Cross-Modal Anomaly Detection

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Pith's one-line read A deep learning framework, CMAD, detects cross-modal anomalies by learning a consensus feature space where consistent views of an instance are pulled together and mismatched views are pushed apart.

desk verdict CMAD is a clean metric-learning formulation for cross-modal label-mismatch detection, but the evaluation is built on the same rule used to train it, so the headline accuracy numbers are a proof-of-concept at most. read the letter →

arxiv 1908.03848 v1 pith:4OIBBNBL submitted 2019-08-11 cs.LG stat.ML

classification cs.LGstat.ML
keywords cross-modalanomalydetectionmulti-modallearningdeepneuralnetworksconsensusfeaturespacenegativesamplingcosinesimilaritycanonicalcorrelationanalysis
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

Most anomaly detection looks at a single data source. This paper studies instances that are normal within each source but inconsistent across sources—for example, a bank transaction whose user profile, image, and signature disagree. The authors propose CMAD, which trains deep neural networks to map each modality into a shared space, using a pull term for same-label cross-modal pairs and a push term for different-label pairs. At test time, an instance is flagged as an anomaly if the cosine similarity between its two modality embeddings falls below a threshold. On image-plus-text MNIST and RGB-D (color plus depth) data, CMAD reaches 0.9921 and 0.9512 accuracy, exceeding linear CCA, kernel CCA, PLS, HOAD, and an embedding network.

What carries the argument

The central machinery is the consensus feature space built by two deep encoders (a CNN for images and a fully connected network for text), trained with the pull-push objective in Eq. (8). The pull term minimizes $1 - \cos(f(M_A^i), g(M_B^j))$ for positive pairs (same class label), while the push term applies the hinge penalty $\max(0, \cos(\cdot) - \gamma)$ for negative sampled pairs (different labels). At test time, Algorithm 2 computes the cosine similarity $F(M_A^i, M_B^i)$ between the two views of the same instance and flags it as an anomaly if this value is below $\epsilon$. The geometry created by the pull-push loss is what makes the single-threshold test work.

What would settle it

Take a dataset where a human annotator labels genuine cross-modal inconsistencies that are not class-label mismatches—for example, two images of the same object that disagree in style, or transactions where profile and behavior conflict—and check whether CMAD's single-threshold cosine-similarity test separates them from normal instances; if precision and recall stay near chance at every threshold, the framework's core mechanism would be refuted.

Watch

Extended reading notes

Core claim

The central claim is that cross-modal anomalies can be detected by a threshold test in a learned consensus space, provided the space is trained with a pull-push objective that uses class labels to define consistency. The authors define an instance as anomalous when $F(M_A^i, M_B^i) < \epsilon$, where $F$ is cosine similarity between deep embeddings of its two modalities. To learn the embedding, they minimize for positive pairs (same class labels) the loss $1 - F$ and for negative sampled pairs (different labels) a hinge loss $\max(0, F - \gamma)$. This geometry—consistent views close, inconsistent views far—is what makes the single-threshold test work. The experiments on MNIST and RGB-D support the claim that deep nonlinear encoders capture these cross-modal correlations better than linear CCA-style projections or simpler embedding networks.

Load-bearing premise

The training signal and the test definition both assume that cross-modal anomaly equals a class-label mismatch between modalities, and without class labels for all modalities the method cannot be trained.

Editorial extensions

If this is right

  • Anomalies that are invisible when each modality is examined alone become detectable by comparing the same instance across modalities in the learned consensus space.
  • Deep nonlinear encoders are necessary: linear CCA, kernel CCA, PLS, and HOAD all underperform CMAD on both datasets, indicating that shallow correlation models miss the complex cross-modal structure.
  • The pull-push objective extends to more than two modalities by adding pairwise loss terms, a direct extension the paper states.
  • The threshold-based test in Algorithm 2 is computationally cheap, making real-time screening of cross-modal inconsistencies feasible in applications like fraud detection.
  • The method requires class labels for every modality during training; without labels, the positive and negative pair sets cannot be constructed, so the approach is supervised.

Reading between the lines

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

  • The reported evaluations inject anomalies as exactly class-label mismatches via negative sampling, so the accuracy figures measure detection of that particular synthetic inconsistency; real-world cross-modal anomalies that are not simple label disagreements remain untested, and the method's transfer to them is an open question.
  • When labels are scarce, one could replace the class-label pairing with self-supervised signals such as cross-modal nearest-neighbor consistency or augmented-view agreement, though the paper does not explore this direction.
  • The pull-push objective is a form of metric learning; adopting triplet mining or adaptive margin scheduling could improve sample efficiency and stability on larger or noisier datasets.
  • Because the anomaly decision is a single threshold on cosine similarity, calibrating epsilon per modality pair (for instance, image-to-image versus image-to-text) may be needed in practice; the paper tunes epsilon per dataset but leaves automatic calibration unaddressed.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes CMAD, a deep structured framework for cross-modal anomaly detection. The method projects each modality through a deep network into a shared latent space and trains with a contrastive loss (Eq. (8)) that pulls together cross-modal pairs whose class labels agree and pushes apart pairs whose labels disagree. At test time, an instance is declared anomalous if the cosine similarity between its two modality embeddings falls below a threshold (Algorithm 2). Experiments on MNIST (with synthetic text tags) and RGB-D inject anomalies via negative sampling, i.e., by pairing modalities with different class labels, and report high accuracy, precision, and recall, including 0.9921 accuracy on MNIST and 0.9512 on RGB-D.

Significance. If the empirical claims held, the paper would make a useful contribution by framing cross-modal anomaly detection as a representation-learning problem and applying deep nonlinear mappings, a clear improvement over linear CCA-style baselines. The objective in Eq. (8) is well specified and the algorithm (Algorithms 1 and 2) is implementable. However, the evaluation protocol is the main weakness: anomalies are constructed by the same label-mismatch rule that defines the supervised training signal, making the reported accuracy a measure of how well the model reproduces the label-pairing rule rather than its ability to detect general cross-modal inconsistencies. The lack of statistical validation and test-set hyperparameter tuning further undermines the central empirical claim. The method may be effective for detecting class-label mismatches, but the paper's broader claims about real-world anomalies such as identity fraud are not supported by the current evidence.

major comments (3)
  1. [Section IV-A and Eq. (8)] The test anomalies are generated by exactly the same rule used to define the negative training pairs. Section IV-A states that inconsistent pairs are created by randomly sampling pairs {p,q} with y_p^A != y_q^B, which is the definition of the negative set N in Eq. (8). Since the loss in Eq. (8) explicitly trains F to be low for such pairs, and the anomaly score in Algorithm 2 is F itself, the reported accuracy measures the model's ability to reproduce the label-mismatch rule used to construct both its supervision and its ground truth. This is circular with respect to the claim of general cross-modal anomaly detection: real-world anomalies described in the introduction (e.g., identity fraud where the profile does not match the signature) are not defined solely by class-label disagreement, and the paper provides no evidence that the approach transfers to such settings. An evaluation with anomalies defined independently of the training labels is needed to support the central claim.
  2. [Tables II and III; Section IV-D] All results are reported as single numbers without error bars, multiple runs, or significance tests. The injected anomaly counts are small (1,019 for MNIST, 545 for RGB-D), and deep network training is stochastic, so the reported margins over the Embedding Network baseline cannot be evaluated without variance estimates. More seriously, Section IV-D tunes the threshold epsilon and margin gamma directly on the test set, reporting the best accuracy achieved at epsilon=0.3 and gamma=0.3. Since epsilon is the decision threshold in Algorithm 2, selecting it on the test data likely inflates the reported detection accuracy. The authors should use a validation split for hyperparameter selection and report mean and standard deviation over multiple random seeds and injected anomaly sets.
  3. [Section III-B, Eq. (8) vs. Algorithm 2] The training pairs in S and N are defined between two different instances (i,j) or (p,q), whereas the anomaly score in Algorithm 2 is computed on the two modalities of the same instance, F(M_A^i, M_B^i). The manuscript does not explain why the same-instance similarity is expected to be well calibrated by a training objective that operates on cross-instance pairs. If same-instance modality pairs are not representative of the training pair distribution, the threshold epsilon chosen in Section IV-D may not transfer to the actual detection task. The paper should clarify whether same-instance pairs are included in S or N during training, or otherwise justify the calibration step.
minor comments (4)
  1. [Section III-B, paragraph 4] The sentence describing the deep framework states that 'instance pairs with consistent patterns across different modalities are pushed away while the pairs with inconsistent cross-modal patterns are pulled together.' This is the opposite of the objective in Eq. (8) and of the rest of the paper; it should read 'pulled together' for consistent patterns and 'pushed away' for inconsistent ones.
  2. [Section IV-A, Eq. (1) and Algorithm 2] The definition in Eq. (1) uses the condition F < epsilon for an anomaly, while Algorithm 2 writes 'if F(M_A^i, M_B^i) - epsilon < 0'. These are equivalent but the notational inconsistency should be harmonized.
  3. [Section IV-E, Eq. (12)] Eq. (12) reuses the threshold epsilon as a binarization threshold for weighting neighbor images in the reconstruction case study, but it is unclear whether this is the same epsilon used in Algorithm 2 or a separately tuned value; this should be clarified.
  4. [Table I] The caption contains a typo: 'Hyaperparameters' should be 'Hyperparameters'.

Circularity Check

2 steps flagged · score 6.0 of 10

High accuracy reflects the same label-mismatch rule used for training and test; detection reduces to thresholding the trained objective.

  1. self definitional [Section III-A Definition (Eq. 1), Section III-B (Eq. 8), Algorithm 2]
    "F (Mi A, Mi B)<ϵ, (1) ... we regard that the ith instance is normal across the two modalities; otherwise, the ith instance is a cross-modal anomaly. ... if F (Mi A, Mi B)−ϵ< 0 then the ith instance is a cross-modal anomaly;"

    The anomaly score F is the same cosine similarity that the training loss in Eq. (8) directly minimizes for positive pairs and maximizes for negative pairs. Algorithm 2 then thresholds this same F at test time, so the detection outcome is a direct read-out of the training objective. Since the training loss is defined over exactly the same function used to define anomalies in Eq. (1), the 'prediction' that inconsistent patterns have low F is true by construction rather than by independent inference.

  2. fitted input called prediction [Section IV-A (Datasets) and Section III-B (negative set N in Eq. (8))]
    "To generate the cross-modal anomalies, we adopt a widely used injection method - negative sampling, to create a number of instance pairs such that their patterns are inconsistent across different modalities. ... we randomly sample a number of k instance pairs {p,q} such that their cross-modal patterns are different such that y_p^A ≠ y_q^B. ... N ={(p,q)} be the set of the negative samples where their cross-modal patterns are inconsistent (i.e., y_p^A ≠ y_q^B)."

    The test anomalies are generated by the same label-mismatch rule y_p^A ≠ y_q^B that defines the negative training pairs in Eq. (8). The ground-truth anomalous instances are therefore produced from the same supervision signal the model was explicitly trained to separate. Reported accuracies, including 0.9921 on MNIST and 0.9512 on RGB-D, measure how well the model reproduces this label-pairing rule on held-out pairs, not its ability to detect a broader class of cross-modal inconsistencies such as the bank-fraud or identity-mismatch examples described in the introduction.

full rationale

The paper's central empirical claim is partially circular because the anomaly definition, training objective, detection threshold, and test-injection rule all use the same ingredients. Eq. (1) defines anomalies as instances with low cross-modal similarity F; Eq. (8) trains the network to make F high for same-label pairs and low for different-label pairs; Algorithm 2 thresholds F; and Section IV-A constructs ground-truth anomalies from random pairs with different labels, exactly the negative-sampling rule used for training. Thus the test set does not provide an independent definition of 'cross-modal anomaly' beyond the training signal, and the strong accuracy numbers largely reflect alignment between supervision and ground-truth generation rather than a verified general anomaly-detection capability. The relative comparison against supervised baselines using the same protocol may still be informative as a methodology benchmark, and the self-citations in the references are peripheral rather than load-bearing. However, because the central 'prediction' reduces by construction to the training objective and the test definition, a score of 6 is appropriate.

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

The central claim rests on a supervised training signal (class labels) and a synthetic anomaly definition; the method is a contrastive metric learning approach rather than an unsupervised anomaly detector.

free parameters (5)
  • Margin gamma = 0.3 (best on MNIST)
    Hyperparameter in the push loss, Eq. (8), tuned over {-0.4..0.5} on the test set.
  • Threshold epsilon = 0.3 (best on MNIST)
    Anomaly threshold in Eq. (1), tuned over {-0.2..0.7} on the test set.
  • Lambda (loss weight) = 1
    Set empirically for the push term in Eq. (8).
  • Network layer dimensions = 784-1440-1280-320-150-50 for MNIST images
    Architectures chosen by hand for each modality and dataset, Table I.
  • Number of injected anomalies = 1019 (MNIST), 545 (RGB-D)
    Dataset construction choice; affects test balance and difficulty.
assumptions (4)
  • domain assumption Class labels y_i^k are available for every instance in every modality during training.
    Positive pairs S and negative pairs N in Eq. (5)-(8) are defined by comparing labels; without labels the method cannot train.
  • domain assumption Anomalies in the test set are instances whose two modalities have different class labels.
    This is the injection procedure in Section IV-A; the model is evaluated only on this definition.
  • domain assumption A deep neural network with enough capacity can learn a consensus space where cosine similarity reflects label agreement.
    Implicit in the use of CNNs and fully connected nets; no theoretical guarantee is provided.
  • domain assumption Cosine similarity is an appropriate measure of cross-modal consistency.
    Chosen in Eq. (4) without comparison to alternative distances.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep Structured Cross-Modal Anomaly Detection." pith.science (2026). https://pith.science/paper/4OIBBNBL

@misc{pith2026190803848,
  author       = {Pith},
  title        = {Pith review of: Deep Structured Cross-Modal Anomaly Detection},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4OIBBNBL}},
  note         = {Machine review of arXiv:1908.03848}
}
read the original abstract

Anomaly detection is a fundamental problem in data mining field with many real-world applications. A vast majority of existing anomaly detection methods predominately focused on data collected from a single source. In real-world applications, instances often have multiple types of features, such as images (ID photos, finger prints) and texts (bank transaction histories, user online social media posts), resulting in the so-called multi-modal data. In this paper, we focus on identifying anomalies whose patterns are disparate across different modalities, i.e., cross-modal anomalies. Some of the data instances within a multi-modal context are often not anomalous when they are viewed separately in each individual modality, but contains inconsistent patterns when multiple sources are jointly considered. The existence of multi-modal data in many real-world scenarios brings both opportunities and challenges to the canonical task of anomaly detection. On the one hand, in multi-modal data, information of different modalities may complement each other in improving the detection performance. On the other hand, complicated distributions across different modalities call for a principled framework to characterize their inherent and complex correlations, which is often difficult to capture with conventional linear models. To this end, we propose a novel deep structured anomaly detection framework to identify the cross-modal anomalies embedded in the data. Experiments on real-world datasets demonstrate the effectiveness of the proposed framework comparing with the state-of-the-art.

Figures

Figures reproduced from arXiv: 1908.03848 by the authors.

Figure 1
Figure 1. An illustration of the proposed cross-modal anomaly detection framework - CMAD. In Figure 1(a), different colors represent different [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Precision (left) and recall (Right) in the MNIST dataset with [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Left: the distribution of images in the projected space which [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Reconstruction result in MNIST. According to finding the [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 24 canonical work pages

  1. [1]

    Anomaly detection: A survey,

    V . Chandola, A. Banerjee, and V . Kumar, “Anomaly detection: A survey,” ACM computing surveys (CSUR) , vol. 41, no. 3, p. 15, 2009

  2. [2]

    Muvir: Multi-view rare category detection

    D. Zhou, J. He, K. S. Candan, and H. Davulcu, “Muvir: Multi-view rare category detection.” in IJCAI, 2015, pp. 4098–4104

  3. [3]

    A spectral framework for detecting inconsistency across multi-source object re- lationships,

    J. Gao, W. Fan, D. Turaga, S. Parthasarathy, and J. Han, “A spectral framework for detecting inconsistency across multi-source object re- lationships,” in Data Mining (ICDM), 2011 IEEE 11th International Conference on. IEEE, 2011, pp. 1050–1055

  4. [4]

    Canonical correlation analysis

    B. Thompson, “Canonical correlation analysis.” 2000

  5. [5]

    Kernel and nonlinear canonical correlation analysis,

    P. L. Lai and C. Fyfe, “Kernel and nonlinear canonical correlation analysis,” International Journal of Neural Systems , vol. 10, no. 05, pp. 365–377, 2000

  6. [6]

    Multi-view low-rank analysis for outlier detection,

    S. Li, M. Shao, and Y . Fu, “Multi-view low-rank analysis for outlier detection,” in Proceedings of the 2015 SIAM International Conference on Data Mining . SIAM, 2015, pp. 748–756

  7. [7]

    Collaborative multi-view denoising,

    L. Zhang, S. Wang, X. Zhang, Y . Wang, B. Li, D. Shen, and S. Ji, “Collaborative multi-view denoising,” in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. ACM, 2016, pp. 2045–2054

  8. [8]

    Neural fraud de- tection in credit card operations,

    J. R. Dorronsoro, F. Ginel, C. Sgnchez, and C. Cruz, “Neural fraud de- tection in credit card operations,” IEEE transactions on neural networks, vol. 8, no. 4, pp. 827–834, 1997

Show all 32 references
  1. [9]

    A survey of data mining and machine learning methods for cyber security intrusion detection,

    A. L. Buczak and E. Guven, “A survey of data mining and machine learning methods for cyber security intrusion detection,” IEEE Commu- nications Surveys & Tutorials , vol. 18, no. 2, pp. 1153–1176, 2016

  2. [10]

    A survey on wearable sensor- based systems for health monitoring and prognosis,

    A. Pantelopoulos and N. G. Bourbakis, “A survey on wearable sensor- based systems for health monitoring and prognosis,” IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews) , vol. 40, no. 1, pp. 1–12, 2010

  3. [11]

    Anomaly detection and classification for hyperspectral imagery,

    C.-I. Chang and S.-S. Chiang, “Anomaly detection and classification for hyperspectral imagery,” IEEE transactions on geoscience and remote sensing, vol. 40, no. 6, pp. 1314–1325, 2002

  4. [12]

    Combining negative selection and classification techniques for anomaly detection,

    F. Gonzalez, D. Dasgupta, and R. Kozma, “Combining negative selection and classification techniques for anomaly detection,” in Evolutionary Computation, 2002. CEC’02. Proceedings of the 2002 Congress on , vol. 1. IEEE, 2002, pp. 705–710

  5. [13]

    Intrusion detection with unlabeled data using clustering,

    L. Portnoy, E. Eskin, and S. Stolfo, “Intrusion detection with unlabeled data using clustering,” in In Proceedings of ACM CSS Workshop on Data Mining Applied to Security (DMSA-2001 . Citeseer, 2001

  6. [14]

    Specae: Spectral autoen- coder for anomaly detection in attributed networks,

    Y . Li, X. Huang, J. Li, M. Du, and N. Zou, “Specae: Spectral autoen- coder for anomaly detection in attributed networks,” 2019

  7. [15]

    Clustering- based anomaly detection in multi-view data,

    A. Marcos Alvarez, M. Yamada, A. Kimura, and T. Iwata, “Clustering- based anomaly detection in multi-view data,” in Proceedings of the 22nd ACM international conference on Conference on information & knowledge management. ACM, 2013, pp. 1545–1548

  8. [16]

    Exploiting similarities among languages for machine translation,

    T. Mikolov, Q. V . Le, and I. Sutskever, “Exploiting similarities among languages for machine translation,” arXiv preprint arXiv:1309.4168 , 2013

  9. [17]

    Multimodal deep learning,

    J. Ngiam, A. Khosla, M. Kim, J. Nam, H. Lee, and A. Y . Ng, “Multimodal deep learning,” in Proceedings of the 28th international conference on machine learning (ICML-11) , 2011, pp. 689–696

  10. [18]

    Efficient learning of deep boltz- mann machines,

    R. Salakhutdinov and H. Larochelle, “Efficient learning of deep boltz- mann machines,” in Proceedings of the thirteenth international confer- ence on artificial intelligence and statistics , 2010, pp. 693–700

  11. [19]

    Speech recognition with deep recurrent neural networks,

    A. Graves, A.-r. Mohamed, and G. Hinton, “Speech recognition with deep recurrent neural networks,” in Acoustics, speech and signal pro- cessing (icassp), 2013 ieee international conference on . IEEE, 2013, pp. 6645–6649

  12. [20]

    Graph recurrent networks with attributed random walks,

    X. Huang, Q. Song, Y . Li, and X. Hu, “Graph recurrent networks with attributed random walks,” 2019

  13. [21]

    Is a single vector enough? exploring node polysemy for network embedding,

    N. Liu, Q. Tan, Y . Li, H. Yang, J. Zhou, and X. Hu, “Is a single vector enough? exploring node polysemy for network embedding,”arXiv preprint arXiv:1905.10668, 2019

  14. [22]

    Dropout: a simple way to prevent neural networks from over- fitting,

    N. Srivastava, G. Hinton, A. Krizhevsky, I. Sutskever, and R. Salakhut- dinov, “Dropout: a simple way to prevent neural networks from over- fitting,” The Journal of Machine Learning Research , vol. 15, no. 1, pp. 1929–1958, 2014

  15. [23]

    MNIST handwritten digit database,

    Y . LeCun and C. Cortes, “MNIST handwritten digit database,” 2010. [Online]. Available: http://yann.lecun.com/exdb/mnist/

  16. [24]

    Distributed representations of words and phrases and their composi- tionality,

    T. Mikolov, I. Sutskever, K. Chen, G. S. Corrado, and J. Dean, “Distributed representations of words and phrases and their composi- tionality,” in Advances in neural information processing systems , 2013, pp. 3111–3119

  17. [25]

    Glove: Global vectors for word representation,

    J. Pennington, R. Socher, and C. Manning, “Glove: Global vectors for word representation,” in Proceedings of the 2014 conference on empirical methods in natural language processing (EMNLP) , 2014, pp. 1532–1543

  18. [26]

    A large-scale hierarchical multi- view rgb-d object dataset,

    K. Lai, L. Bo, X. Ren, and D. Fox, “A large-scale hierarchical multi- view rgb-d object dataset,” in Robotics and Automation (ICRA), 2011 IEEE International Conference on . IEEE, 2011, pp. 1817–1824

  19. [27]

    A kernel method for canonical correlation analysis,

    S. Akaho, “A kernel method for canonical correlation analysis,” arXiv preprint cs/0609071, 2006

  20. [28]

    Partial least square regression (pls regression),

    H. Abdi, “Partial least square regression (pls regression),” Encyclopedia for research methods for the social sciences , vol. 6, no. 4, pp. 792–795, 2003

  21. [29]

    Learning two-branch neural networks for image-text matching tasks,

    L. Wang, Y . Li, J. Huang, and S. Lazebnik, “Learning two-branch neural networks for image-text matching tasks,” IEEE Transactions on Pattern Analysis and Machine Intelligence , 2018

  22. [30]

    Learning deep structure-preserving image-text embeddings,

    L. Wang, Y . Li, and S. Lazebnik, “Learning deep structure-preserving image-text embeddings,” in Proceedings of the IEEE conference on computer vision and pattern recognition , 2016, pp. 5005–5013

  23. [31]

    Deep structured energy based models for anomaly detection,

    S. Zhai, Y . Cheng, W. Lu, and Z. Zhang, “Deep structured energy based models for anomaly detection,” arXiv preprint arXiv:1605.07717, 2016

  24. [32]

    Heterogeneous network embedding via deep architectures,

    S. Chang, W. Han, J. Tang, G.-J. Qi, C. C. Aggarwal, and T. S. Huang, “Heterogeneous network embedding via deep architectures,” in Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining . ACM, 2015, pp. 119–128

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.