Pith. sign in

REVIEW 4 major objections 5 minor 43 references

Spatio-temporal Latent Representations for the Analysis of Acoustic Scenes in-the-wild

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a self-supervised variational autoencoder trained on sparse acoustic-event embeddings can regroup unlabeled in-the-wild audio by spatio-temporal acoustic scene, separating indoor locations from metro locations where…

desk verdict Honest, clearly written exploratory pipeline, but the advertised VAE separation of indoor vs metro audio is one user's eyeballed heatmap colored by labels the authors themselves say cannot support any metric. read the letter →

arxiv 2412.07648 v1 pith:SQWXQQTH submitted 2024-12-10 eess.AS

Claudia Montero-Ramírez , Esther Rituerto-González , Carmen Peláez-Moreno This is my paper · ORCID
classification eess.AS
keywords acousticsceneanalysisself-supervisedlearningvariationalautoencoderlatentspacein-the-wildaudioeventdetectionTF-IDFNode2Vec
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that a variational autoencoder can turn unlabeled, in-the-wild audio into a latent space where acoustic scenes separate by their spatial context. Using one-minute recordings from a wearable dataset, the authors convert detected acoustic events into TF-IDF and Node2Vec embeddings, then train a per-user VAE on those sparse vectors. They report that in the latent space, the audio segments form clusters that line up with indoor locations (home and work) and metro locations, while the raw embeddings show no such grouping. If this holds, acoustic scenes can be characterized without clean location labels, which is a step toward detecting anomalous or violent situations from sound in real time.

What carries the argument

The load-bearing object is the per-user variational autoencoder and the embedding pipeline that feeds it. Audio is cut into one-minute segments, passed through YAMNet to get probabilities for 521 AudioSet event classes, binarized at a 99th-percentile threshold, and converted to a concatenated TF-IDF plus Node2Vec vector per segment. The VAE compresses these sparse vectors into a continuous latent space; cosine distance in that space and t-SNE visualizations are used to expose the cluster structure. Training uses stochastic gradient descent with momentum and an exponentially decaying learning rate to cope with the sparsity of the input matrices.

What would settle it

Compute the cosine-distance matrix within and between location groups in the VAE latent space for all users and compare it to the same matrix computed after randomly shuffling the location labels; if shuffled labels produce clusters as separated as the real ones, the grouping is not driven by acoustic scene content. A second check: re-run the t-SNE visualization with several random seeds and see whether the indoor/metro split persists, since t-SNE initialization is nondeterministic.

Watch

Extended reading notes

Core claim

The central discovery is that the VAE's latent space is more discriminative for acoustic scenes than the raw embeddings. For one user, the cosine-distance heatmap of latent vectors shows two clear clusters: one for indoor locations (home and work) and one for metro locations, whereas the heatmap of the input TF-IDF/Node2Vec embeddings is comparatively uniform. t-SNE projections over training epochs show the latent points gradually regrouping from a random scatter into location-associated clusters. The authors frame this as self-supervised spatio-temporal characterization: the model never sees location labels during training, and the noisy GPS-derived tags serve only as guides for interpreting the clusters.

Load-bearing premise

The load-bearing premise is that the noisy GPS-derived and self-reported location tags are accurate enough that the latent clusters can be interpreted as 'indoor' and 'metro' acoustic scenes; the paper itself says GPS is often inaccurate and that any metric using these labels would not be significant.

Editorial extensions

If this is right

  • Acoustic scenes in the wild can be characterized without accurate location labels, since the VAE organizes the data before any label is used.
  • The latent space gives a more separable representation than the count-based and ontology-based embeddings alone, at least for the demonstrated user.
  • The same self-supervised pipeline could be applied to other users or datasets to obtain personalized scene models, and the per-user setup leaves room for personalization.
  • The authors position this characterization as the first step toward detecting anomalous or violent acoustic contexts in real time, with zero-shot extensions named as future work.

Reading between the lines

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

  • A quantitative cluster-quality test (e.g., silhouette score computed on the latent vectors per user) would let the claimed indoor/metro separation be checked without relying on t-SNE's random initialization, which the paper itself flags as inconclusive.
  • If the effect generalizes across the 13 analyzed users, the same embeddings could serve as a pretrained representation for downstream audio-context classifiers, reducing the need for labeled risk-scene data.
  • The dependence of the clusters on YAMNet's 521 event classes could be probed by swapping in other AudioSet-based embeddings (e.g., PANNs or VGGish) and checking whether the same indoor/metro structure persists.
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 a self-supervised pipeline for characterizing acoustic scenes in-the-wild from the WE-LIVE dataset. Audio is converted to discrete acoustic event activations using YAMNet, then to TF-IDF and Node2Vec embeddings, and finally a per-user linear VAE is trained to obtain a latent space. The authors claim that cosine-distance heatmaps and t-SNE visualizations of this latent space reveal distinct acoustic scenes, in particular separating indoor locations (home, work) from metro locations, whereas the raw embeddings do not. The main evidence is qualitative: one user's heatmap and t-SNE plots, with location pseudo-labels derived from GPS and self-reports.

Significance. If the central claim were properly supported, the paper would offer a useful exploratory result: a self-supervised VAE could impose structure on noisy, in-the-wild audio embeddings in a way that corresponds to meaningful acoustic environments, with potential applications to affective computing and risk detection. The work also contributes a description of the WE-LIVE dataset and a practical comparison of YAMNet and PANNs CNN14 inference times. However, the current evidence is qualitative and based on a single user, and the authors themselves state in Section 6.1 that the pseudo-labels are too unreliable for any metric-based evaluation. The significance is therefore conditional on substantial additional validation.

major comments (4)
  1. [Section 6.4 and Section 5, Fig. 5] The central claim that the VAE latent space 'is able to find a cluster related to metro locations and another cluster for indoor locations' rests on visual inspection of one user's cosine-distance heatmap. Section 6.1 explicitly states that GPS-derived location tags are inaccurate and that 'any outcome linked to an accuracy or any metric using these labels would not be significant,' and Section 5 states that the t-SNE plots 'are not conclusive' due to random initialization. No clustering metric, statistical test, or multi-user replication is provided, so the evidence does not support the central claim as stated. The authors should either supply quantitative validation (e.g., silhouette score, adjusted mutual information, permutation tests against chance structure) on held-out data, or substantially temper the claim to a descriptive observation for an individual user.
  2. [Section 5, Fig. 5] The claim that the VAE latent space is 'more discriminative' than the raw embeddings is supported only by the observation of a 'larger range of similarity metric values.' This is not a valid measure of discriminability, because a VAE can rescale or rotate representations arbitrarily, and a larger range of cosine distances does not imply that the distances separate known acoustic scenes from one another. A quantitative comparison against the raw embeddings, using the same clustering or evaluation protocol on both representations, is needed.
  3. [Section 4.2 and Section 6.1] The embedding pipeline depends on several user-specific and dataset-wide choices: the 99th percentile activation threshold for YAMNet probabilities, the hexagon grid edge of 0.0015, the restriction to the top 10 locations per user, the audio segment length of one minute, and the VAE hyperparameters. These are selected on the same data used for the qualitative evaluation, and no sensitivity analysis is reported. Because the activation threshold directly controls which acoustic events enter the TF-IDF/Node2Vec representation, the robustness of the claimed clustering to these choices should be demonstrated.
  4. [Section 6.2] The comparison between YAMNet and PANNs CNN14 is based solely on inference time, yet the paper later uses YAMNet outputs as the basis for the acoustic-scene representation. The choice of event detector could affect the resulting embeddings and the observed latent structure; without any evaluation of representation quality or downstream clustering behavior, the choice of YAMNet is not well justified for the paper's central goal.
minor comments (5)
  1. [Section 5 and Section 6.4] Section 5 says the heatmap is shown 'for one user' and the t-SNE evolution 'for one user,' while Section 6.4 refers to 'the heatmap related to one user location (Fig. 5)' and 'another user' for Fig. 6; the text should clarify whether Figs. 5 and 6 are for the same user or different users.
  2. [Section 4.4] The VAE architecture is described only as 'linear' with ReLU activations and tanh on the last decoder layer; the number of layers, hidden dimensions, latent dimension, and early-stopping criterion are not specified, which limits reproducibility.
  3. [Throughout] There are several typos and incomplete references: 'recorede' in Section 3, 'Comparision' in Table 1, and references [4], [5], and [41] lack complete bibliographic information; the authors should also replace blog-style citations with primary sources where available.
  4. [Section 4.3] The sentence 'A 5x521 matrix is obtained related to Node2Vec embeddings' is unclear; the text should state explicitly that each of the 521 AudioSet events is represented by a 5-dimensional vector, and that the final embedding for a one-minute audio is a 6x521 matrix.
  5. [Section 3] The description of the WE-LIVE dataset mentions 14 female volunteers and that one was discarded due to insufficient audio data, but the number of users actually used in the experiments is not stated consistently; please state explicitly that 13 users were used in the analysis.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the VAE latent-space analysis is trained without location labels and its cluster interpretation is not enforced by construction.

full rationale

The central claim is that the VAE latent space regroups acoustic scenes, with indoor and metro locations appearing as clusters in a cosine-distance heatmap for one user. The VAE is trained on audio-event embeddings (TF-IDF and Node2Vec from YAMNet outputs) without using GPS or self-reported labels; the location tags enter only after training, as colors for visualization and interpretation. Thus the observed clustering is not forced to match the pseudo-labels by the training objective. The paper itself states that the pseudo-labels are too noisy for supervised learning and that any accuracy metric using them would not be significant (Section 6.1), which is a limitation on evidential strength rather than a circular step. The activation threshold (99th percentile) and hyperparameters are chosen on the same data, but they are not fitted to the location labels or to the cluster claim, so this is standard modeling practice rather than a fitted-input-called-prediction pattern. The self-citations (e.g., [26] for the in-house WE-LIVE dataset, [19] for cosine-distance usage) provide dataset description and methodological precedent; they do not supply the latent-space separation result. No equation in the paper defines the VAE representation in terms of the location labels, and no prediction is derived from a parameter fitted to the target quantity. The evidence is qualitative and limited to one user, but that is a correctness/robustness concern, not circularity. Overall, the derivation chain is self-contained with respect to the paper's inputs.

Assumptions & free parameters 7 free parameters · 5 assumptions · 0 invented entities

The central claim rests on two families of assumptions: that the audio preprocessing chain (YAMNet threshold, TF-IDF, Node2Vec, VAE) preserves scene-discriminative structure, and that the noisy location labels are accurate enough to name the observed clusters. No new theoretical entity is introduced; all free parameters are either hand-chosen or fitted per user on the same data. The VAE itself is not circular with respect to the labels, but the interpretation is label-dependent.

free parameters (7)
  • YAMNet activation threshold = 99th percentile per user; about 5 events/second
    All acoustic event probabilities are binarized using a data-dependent threshold chosen per user; this determines which events enter the TF-IDF and Node2Vec embeddings and therefore shapes the latent clustering.
  • Hexagon grid edge = 0.0015 degrees (about 60 m radius)
    GPS coordinates are discretized to cells; the cell size is chosen by hand and common to all users; it directly defines the location pseudo-labels.
  • Number of top locations per user = 10
    Only the 10 most frequent hexagons per user are kept; this filters the audio set and restricts which locations appear in the analysis.
  • Audio segment length = 60 seconds
    Audios are cut to one minute as the document unit; this length controls how many acoustic events are aggregated per embedding.
  • Node2Vec embedding dimension = 5
    Each AudioSet event is represented by a 5-dimensional vector; no justification is given for this dimensionality.
  • VAE learning hyperparameters = alpha0=1e-5, gamma=0.99, beta=0.9
    Learning rate schedule and momentum are fixed by hand; the VAE training is per-user with early stopping, so final parameters are data-dependent.
  • VAE latent dimension = not reported
    Not stated; a free design choice that determines the geometry of the latent space.
assumptions (5)
  • domain assumption YAMNet probabilities are reliable indicators of acoustic events in unconstrained in-the-wild audio
    Section 4.2 treats YAMNet output as ground-truth event presence after thresholding; in-the-wild recording conditions may bias these probabilities, as the authors acknowledge in Section 6.1.
  • domain assumption The AudioSet ontology captures semantic relationships between sounds that are useful for scene characterization
    Section 4.3 uses the ontology graph to compute Node2Vec embeddings; if ontology edges do not correspond to acoustic co-occurrence in real scenes, the embeddings may be uninformative.
  • ad hoc to paper A linear VAE trained on flattened TF-IDF/Node2Vec embeddings produces a latent space in which Euclidean/cosine distances reflect acoustic scene similarity
    Section 4.4 introduces the VAE as the core mechanism; no evidence is given that the reconstruction objective induces scene-discriminative structure rather than merely compressing the input.
  • domain assumption GPS-derived location tags and self-reported situational labels are accurate enough to label clusters in visualizations
    Section 4.1 defines pseudo-labels from sparse GPS cells; Section 6.1 states GPS is inaccurate and labels cannot support supervised metrics. The interpretation of clusters as 'indoor' or 'metro' depends on these labels.
  • domain assumption t-SNE projections are representative of the latent geometry despite random initialization
    Section 5 uses t-SNE to claim regrouping over epochs, but Section 6.3 admits t-SNE initialization is random and plots are not conclusive.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Spatio-temporal Latent Representations for the Analysis of Acoustic Scenes in-the-wild." pith.science (2026). https://pith.science/paper/SQWXQQTH

@misc{pith2026241207648,
  author       = {Pith},
  title        = {Pith review of: Spatio-temporal Latent Representations for the Analysis of Acoustic Scenes in-the-wild},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SQWXQQTH}},
  note         = {Machine review of arXiv:2412.07648}
}
read the original abstract

In the field of acoustic scene analysis, this paper presents a novel approach to find spatio-temporal latent representations from in-the-wild audio data. By using WE-LIVE, an in-house collected dataset that includes audio recordings in diverse real-world environments together with sparse GPS coordinates, self-annotated emotional and situational labels, we tackle the challenging task of associating each audio segment with its corresponding location as a pretext task, with the final aim of acoustically detecting violent (anomalous) contexts, left as further work. By generating acoustic embeddings and using the self-supervised learning paradigm, we aim to use the model-generated latent space to acoustically characterize the spatio-temporal context. We use YAMNet, an acoustic events classifier trained in AudioSet to temporally locate and identify acoustic events in WE-LIVE. In order to transform the discrete acoustic events into embeddings, we compare the information-retrieval-based TF-IDF algorithm and Node2Vec as an analogy to Natural Language Processing techniques. A VAE is then trained to provide a further adapted latent space. The analysis was carried out by measuring the cosine distance and visualizing data distribution via t-Distributed Stochastic Neighbor Embedding, revealing distinct acoustic scenes. Specifically, we discern variations between indoor and subway environments. Notably, these distinctions emerge within the latent space of the VAE, a stark contrast to the random distribution of data points before encoding. In summary, our research contributes a pioneering approach for extracting spatio-temporal latent representations from in-the-wild audio data.

Figures

Figures reproduced from arXiv: 2412.07648 by the authors.

Figure 1
Figure 1. Map of top 10 most frequented cells by a volunteer. [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Final embedding matrix representation for each [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Linear VAE structure. criteria. The following considerations were applied for each user: • A comparison of the cosine distance (Eq. (4), [19]) related to the GPS cell location tags, between the input embeddings (Section 4.3) and the latent space of the VAE (Section 4.4). A cosine distance matrix was computed to compare the input embeddings with each other, as well as another matrix for comparing the vectors within t… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: VAE loss function evolution during epochs example [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: t-SNE algorithm visualization during training for one [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 5
Figure 5. Figure 5: Cosine distance between pseudo-labeled vector [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 33 canonical work pages

  1. [1]

    Rituerto-Gonz ´alez, C

    E. Rituerto-Gonz ´alez, C. Luis-Mingueza, and C. Pel ´alez-Moreno, ”Using Audio Events to Extend a Multi-modal Public Speaking Database with Reinterpreted Emotional Annotations,” in IberSPEECH 2021, pp. 61–65, 2021. [Online]. Available: https://doi.org/10.2143/iberspeech. 2021-13

    work page doi:10.2143/iberspeech 2021

  2. [2]

    Eerola and J

    T. Eerola and J. K. Vuoskoski, ”The influence of acoustic context on emotional responses to sounds,” Psychology of Music, vol. 39, no. 3, pp. 335-355, 2011. doi: 10.1177/0305735610377905

    work page doi:10.1177/0305735610377905 2011

  3. [3]

    Doersch, S

    C. Doersch, S. Singh, A. Gupta, J. Sivic, and A. A. Efros, ”What Makes Paris Look like Paris?” in Proceedings of the 2012 European Conference on Computer Vision, 2012

    work page 2012

  4. [4]

    J ¨arvel¨ainen and V

    H. J ¨arvel¨ainen and V . V ¨alim¨aki, ”Soundtracks datasets for music and emotion,” [Online]. Available: https://www.jyu.fi/hytk/fi/laitokset/mutku/ en/research/projects2/past-projects/coe/materials/emotion/soundtracks/ Index. Retrieved on September 11, 2023

    work page 2023

  5. [5]

    Carvalho and J

    S. Carvalho and J. Leite, ”The Emotional Movie Database (EMDB): A Self-Report and Psychophysiological Study,” [Online]. Retrieved on September 11, 2023

    work page 2023

  6. [6]

    Plakal and D

    M. Plakal and D. Ellis, ”Yamnet,” Jan 2020. [Online]. Available: https: //github.com/tensorflow/models/tree/master/research/audioset/yamnet

    work page 2020

  7. [7]

    Q. Kong, Y . Cao, T. Iqbal, Y . Wang, W. Wang, and M. D. Plumbley, ”PANNs: Large-Scale Pretrained Audio Neural Networks for Audio Pattern Recognition,” arXiv preprint, 2019. [Online]. Available: http: //arxiv.org/abs/1912.10211

    arXiv 2019

  8. [8]

    Sparck Jones, ”A Statistical Interpretation of Term Specificity and Its Application in Retrieval,” Journal of Documentation, vol

    K. Sparck Jones, ”A Statistical Interpretation of Term Specificity and Its Application in Retrieval,” Journal of Documentation, vol. 28, pp. 11-21,

    work page

Show all 43 references
  1. [9]

    Grover and J

    A. Grover and J. Leskovec, ”Node2vec: Scalable feature learning for networks,” in Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 2016, pp. 855–864. [Online]. Available: https://doi.org/10.1145/2939672.2939754

    2016

  2. [10]

    [Online]

    Nikhil Birajdar, ”Word2Vec Research Paper Explained,” Towards Data Science, March 2021. [Online]. Available: https://towardsdatascience. com/word2vec-research-paper-explained-205cb7eecc30

    2021

  3. [11]

    Gatopoulos and J

    I. Gatopoulos and J. M. Tomczak, ”Self-Supervised Variational Auto-Encoders,” Entropy, vol. 23, no. 6, p. 747, 2021. [Online]. Available: https://doi.org/10.3390/e23060747

    2021 doi

  4. [12]

    Zhou, ”V AE: Variational Autoencoders — How to Employ Neural Networks to Generate New Images,” Towards Data Science, 2023

    V . Zhou, ”V AE: Variational Autoencoders — How to Employ Neural Networks to Generate New Images,” Towards Data Science, 2023. [Online]. Available: [Accessed: Aug. 8, 2023]

    2023

  5. [14]

    J. A. Miranda, E. Rituerto-Gonz ´alez, L. Guti ´errez-Mart´ın, C. Luis-Mingueza, M. F. Canabal, A. R. B ´arcenas, J. M. Lanza-Guti ´errez, C. Pel ´aez-Moreno, y C. L ´opez-Ongil, ”WEMAC: Women and Emotion Multi-modal Affective Computing dataset,” arXiv preprint, Mar. 2022. [On...

    2022 arXiv

  6. [15]

    Abeßer, ”A review of deep learning based methods for acoustic scene classification,” Applied Sciences (Switzerland), vol

    J. Abeßer, ”A review of deep learning based methods for acoustic scene classification,” Applied Sciences (Switzerland), vol. 10, no. 6, 2020. [Online]. Available: https://doi.org/10.3390/app10062020

    2020 doi

  7. [16]

    S. S. Kushwaha and M. Fuentes, ”A Multimodal Prototypical Approach for Unsupervised Sound Classification,” arXiv preprint, 2023. [Online]. Available: http://arxiv.org/abs/2306.12300

    2023 arXiv

  8. [17]

    Barchiesi, D

    D. Barchiesi, D. Giannoulis, D. Stowell, and M. D. Plumbley, ”Acoustic Scene Classification,” IEEE Signal Processing Magazine, vol. 31, no. 3, pp. 100-114, 2014. [Online]. Available: https://doi.org/10.1109/MSP. 2014.2326181

    2014

  9. [18]

    Heittola, A

    T. Heittola, A. Mesaros, A. Eronen, and T. Virtanen, ”AUDIO CONTEXT RECOGNITION USING AUDIO EVENT HISTOGRAMS,” 2010

    2010

  10. [19]

    Luis-Mingueza, E

    C. Luis-Mingueza, E. Rituerto-Gonz ´alez, and C. Pel ´aez-Moreno, ”Bridging the Semantic Gap with Affective Acoustic Scene Analysis: an Information Retrieval-based Approach,” in IberSPEECH 2022, pp. 91–95, 2022. [Online]. Available: https://doi.org/10.21437/iberspeech. 2022-19

    2022 doi

  11. [20]

    Latif, M

    S. Latif, M. Usama, M. I. Malik, and B. W. Schuller, ”Can Large Language Models Aid in Annotating Speech Emotional Data? Uncovering New Frontiers,” arXiv preprint, 2023. [Online]. Available: http://arxiv.org/abs/2307.06090 8

    2023 arXiv

  12. [21]

    J. F. Gemmeke, D. P. W. Ellis, D. Freedman, A. Jansen, W. Lawrence, R. C. Moore, M. Plakal, and M. Ritter, ”AUDIO SET: AN ONTOLOGY AND HUMAN-LABELED DATASET FOR AUDIO EVENTS,” Google, Inc., Mountain View, CA, and New York, NY , USA, 2017

    2017

  13. [22]

    Hershey, S

    M. Hershey, S. Chaudhuri, D. P. W. Ellis, J. F. Gemmeke, A. Jansen, R. C. Moore, M. Plakal, D. Platt, R. A. Saurous, B. Seybold, M. Slaney, R. Weiss, and K. Wilson, ”CNN Architectures for Large-Scale Audio Classification,” GitHub repository, TensorFlow,

  14. [23]

    Y . Deng, M. Yang, y R. Zhou, ”A new standardized emotional film database for Asian culture,” Frontiers in Psychology, vol. 8, 2017. [Online]. Available: https://doi.org/10.3389/fpsyg.2017.01941

    2017

  15. [24]

    Blanco-Ruiz, C

    M. Blanco-Ruiz, C. Sainz-De-baranda, L. Guti ´errez-Mart´ın, E. Romero-Perales, y C. L ´opez-Ongil, ”Emotion elicitation under audiovisual stimuli reception: Should artificial intelligence consider the gender perspective?” International Journal of Environmental Research and Pu...

    2020 doi

  16. [25]

    Delegaci ´on de Gobierno contra la Violencia ge G ´enero, Ministerio de Igualdad, Gobierno de Espa ˜na

    Executive Summary of the 2019 Macro-urvey on Violence Against Women 2021. Delegaci ´on de Gobierno contra la Violencia ge G ´enero, Ministerio de Igualdad, Gobierno de Espa ˜na. Available: https://violenciagenero.igualdad.gob.es/violenciaEnCifras/ macroencuesta2015/pdf/RE Macr...

    2019

  17. [26]

    Rituerto Gonz ´alez, ”Multimodal Affective Computing in Wearable Devices with Applications in the Detection of Gender-based Violence,” Doctoral Dissertation, Dept

    E. Rituerto Gonz ´alez, ”Multimodal Affective Computing in Wearable Devices with Applications in the Detection of Gender-based Violence,” Doctoral Dissertation, Dept. of Signal Theory and Communications, Universidad Carlos III de Madrid, Madrid, Spain, 2023

    2023

  18. [27]

    J. A. Miranda Calero, E. Rituerto-Gonz ´alez, C. Luis-Mingueza, M. F. Canabal, A. Ram ´ırez B ´arcenas, J. M. Lanza-Guti ´errez, C. Pel´aez-Moreno, y C. L ´opez-Ongil, ”Bindi: Affective Internet of Things to Combat Gender-Based Violence,” INTERNET OF THINGS JOURNAL, vol. 9, 20...

    2022

  19. [28]

    [Online]

    Capital One, ”Understanding TF-IDF for Machine Learning,” October 6, 2021. [Online]. Available: https://www.capitalone.com/ tech/machine-learning/understanding-tf-idf/

    2021

  20. [29]

    Arya, ”TF-IDF Defined,” KDnuggets, October 2022

    N. Arya, ”TF-IDF Defined,” KDnuggets, October 2022. [Online]. Available: https://www.kdnuggets.com/2022/10/tfidf-defined.html

    2022

  21. [30]

    H. Zhao, P. Rai, L. Du, W. Buntine, D. Phung, y M. Zhou, ”Variational Autoencoders for Sparse and Overdispersed Discrete Data,” [Online]. Available: https://github.com/ethanhezhao/NBV AE

  22. [31]

    Utyamishev y I

    D. Utyamishev y I. Partin-Vaisband, ”Progressive V AE Training on Highly Sparse and Imbalanced Data,” arXiv preprint, 2019. [Online]. Available: http://arxiv.org/abs/1912.08283

    2019 arXiv

  23. [32]

    [Online]

    PyTorch, ”torch.optim.SGD,” PyTorch Documentation, 2023. [Online]. Available: https://pytorch.org/docs/stable/generated/torch.optim.SGD. html

    2023

  24. [33]

    X. Liu, F. Zhang, Z. Hou, Z. Wang, L. Mian, J. Zhang, y J. Tang, ”Self-supervised Learning: Generative or Contrastive,” IEEE Transactions on Knowledge and Data Engineering, 2020. [Online]. Available: https://doi.org/10.1109/TKDE.2021.3090866

    2020

  25. [34]

    [Online]

    PyTorch, ”torch.optim.lr scheduler.ExponentialLR,” PyTorch Documentation, 2023. [Online]. Available: https://pytorch.org/docs/ stable/generated/torch.optim.lr scheduler.ExponentialLR.html

    2023

  26. [35]

    Z. Pan, W. Liu, Z. Meng, y J. Yin, ”Linear Variational Autoencoder for Top-N Recommendation,” in 2022 7th International Conference on Big Data Analytics, ICBDA 2022, 2022, pp. 296–303. [Online]. Available: https://doi.org/10.1109/ICBDA55095.2022.9760352

    2022

  27. [36]

    van der Maaten y G

    L. van der Maaten y G. Hinton, ”Viualizing data using t-SNE Visualizing Data using t-SNE,” Journal of Machine Learning Research, vol. 9, 2008. [Online]. Available: https://www.researchgate.net/publication/ 228339739

    2008

  28. [37]

    Pitsillos, ”MNIST Variational Autoencoder and t-SNE Visualisation,” May 2020

    N. Pitsillos, ”MNIST Variational Autoencoder and t-SNE Visualisation,” May 2020. [Online]. Available: https://npitsillos.github.io/blog/2020/ mnistvae/

    2020

  29. [38]

    Su ´arez, ”Reducci ´on de la dimensionalidad – TSNE,” May

    D. Su ´arez, ”Reducci ´on de la dimensionalidad – TSNE,” May

  30. [39]

    Available: https://huggingface.co/nlptown/ bert-base-multilingual-uncased-sentiment

    Hugging Face, ”Bert Base Multilingual Uncased Sentiment,” [Online]. Available: https://huggingface.co/nlptown/ bert-base-multilingual-uncased-sentiment

  31. [40]

    X. Mei, C. Meng, H. Liu, Q. Kong, T. Ko, C. Zhao, M. D. Plumbley, Y . Zou, y W. Wang, ”WavCaps: A ChatGPT-Assisted Weakly-Labelled Audio Captioning Dataset for Audio-Language Multimodal Research,” arXiv preprint, 2023. [Online]. Available: http://arxiv.org/abs/2303. 17395

    2023

  32. [41]

    [Online]

    Veryfi, ”What is Zero-Shot Learning?” [Blog post], 2023. [Online]. Available: https://www.veryfi.com/technology/zero-shot-learning/ 9

    2023

  33. [1972]

    Available: http://dx.doi.org/10.1108/eb026526

    [Online]. Available: http://dx.doi.org/10.1108/eb026526

    doi

  34. [2017]

    Available: https://github.com/tensorflow/models/tree/ master/research/audioset/vggish

    [Online]. Available: https://github.com/tensorflow/models/tree/ master/research/audioset/vggish

  35. [2022]

    Available: https://apiumhub.com/es/tech-blog-barcelona/ reduccion-de-dimensionalidad-tsne/

    [Online]. Available: https://apiumhub.com/es/tech-blog-barcelona/ reduccion-de-dimensionalidad-tsne/

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.