Pith. sign in

REVIEW 5 major objections 6 minor 29 references

A Time-Series Data Augmentation Model through Diffusion and Transformer Integration

T0 review · 5 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a diffusion model generating only the first time step, followed by a transformer predicting the remaining steps, produces time-series data that improves sign-language classification from 57.6% to 87.6% when used as…

desk verdict A simple diffusion+transformer augmentation pipeline with a promising accuracy gain, but the 'high-quality generation' claim outruns the evidence because full-sequence realism is never properly measured. read the letter →

arxiv 2505.03790 v1 pith:X74C2M7W submitted 2025-05-01 cs.LG cs.AIeess.SP

Yuren Zhang , Zhongnan Pu , Lei Jing This is my paper · ORCID
classification cs.LGcs.AIeess.SP MSC 68T0768T10
keywords time-seriesdataaugmentationdiffusionmodeltransformersignlanguagerecognitiongloveweightedlossfunctionsensortimeseriesclassificationimprovement
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 two-stage diffusion-plus-transformer generator can create plausible, useful time-series data from a small sensor dataset. The authors argue that because a sign's first frame constrains later frames, it is enough to synthesize new first frames with a modified diffusion model and let a transformer predict the rest. The resulting synthetic samples, when added to a real training set of sign-language glove data, improve classification accuracy from 57.6% to 87.6%, surpassing both no augmentation and traditional augmentation. The evidence is one real dataset plus an indirect evaluation: a classifier trained on real data recognizes most generated samples, and augmented training improves held-out accuracy. A sympathetic reader cares because time-series data is expensive and scarce, and most augmentation work targets images and speech.

What carries the argument

The load-bearing object is the two-stage generation chain: a modified diffusion model with linear layers and a sigmoid output produces new $1 \times 34$ first time-step vectors, and a transformer with a view mask of window size $w$ (tested at 1, 3, and 5) writes the remaining 609 time steps conditioned on that first vector and the class label. A weighted loss, built by fitting a 20-degree polynomial to adjacent-step differences and splitting the timeline at quartiles, counteracts the tendency of MSE to flatten generated trajectories in padded no-change regions. The view mask localizes the transformer's attention to the most recent $w$ positions, and the window size is the main control over realism versus diversity.

What would settle it

Group real sign-language samples whose first-step vectors are close in Euclidean distance and check whether their later trajectories still spread widely; if they do, the premise that the first value determines all subsequent readings is false, and the transformer's continuations cannot represent the full real distribution.

Watch

Extended reading notes

Core claim

The paper claims that a simple linear combination of two existing architectures generates useful time-series data: a diffusion model creates only the initial time-step vector, and a transformer autoregressively predicts the rest. On a 34-channel tactile-glove sign-language dataset, the claim is supported by two observations: a classifier trained on real data classifies the generated samples with up to 94.8% accuracy for window size 5, and adding 250 generated samples to the training set lifts classification accuracy from 57.6% to 87.6% with window size 3, outperforming traditional augmented data at 77.2%. The authors interpret the window-size-3 result as evidence that the generated data contributes useful out-of-distribution diversity in addition to realism.

Load-bearing premise

The entire generative pipeline rests on the assumption that the first time step's readings fully determine the later readings of a sign-language action, so any real mid-action variability not fixed by the first frame is missing from the synthetic data.

Editorial extensions

If this is right

  • If the central claim holds, time-series augmentation reduces to two simpler tasks: synthesize a plausible initial state, then let an autoregressive model generate the rest.
  • Adding the generated window-size-3 data to the real training set improves held-out sign-language classification from 57.6% to 87.6%, a gain of 30.0 points over no augmentation.
  • The generated samples are close enough to real data that a classifier trained on real data recognizes most of them, with the window-size-5 set reaching 94.8% recognition accuracy.
  • The view-mask window size becomes a tunable knob that trades realism against out-of-distribution diversity, with the more diverse size-3 set helping the classifier more.
  • Alternating a weighted loss with standard MSE loss prevents the transformer from collapsing all generated trajectories to a flat line when the dataset contains padded no-change tails.

Reading between the lines

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

  • Beyond the paper: if first-frame conditioning is valid, the same two-stage recipe should transfer to any wearable or kinematic sensor stream where an initial pose largely determines the motion, such as gait or gesture data; a direct test would be to run the pipeline on a second dataset with the same code.
  • Beyond the paper: the comparison is against traditional augmentation, not against full-sequence generative models; the paper's own logic suggests a head-to-head with full-sequence diffusion on this glove dataset would be the next test of whether the simplicity claim converts into a performance edge.
  • Beyond the paper: the weighting scheme is derived from one dataset's padding artifact; on time series without padded flat tails, the quartile-based interval division may add noise rather than help, so the method likely needs re-fitting of the quartile boundaries per dataset.
  • Beyond the paper: because window-size-3 data scored lower on recognizability yet produced the largest accuracy gain, the result hints that out-of-distribution diversity, not fidelity, drives augmentation value; a testable extension would measure the relationship between FID scores and downstream accuracy across window sizes.
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

5 major / 6 minor

Summary. The paper proposes a two-stage generative model for time-series data augmentation. A modified Diffusion model generates 34-dimensional first-time-step sensor vectors, and a Transformer with a learnable "view mask" and a weighted loss function generates the remaining 609 time steps conditioned on the first vector and the class label. The method is evaluated on a self-collected sign-language glove dataset containing 10 signs from 10 subjects. The evaluation consists of (i) FID between generated and real first-step vectors, (ii) a classifier's accuracy on generated full sequences, and (iii) downstream classification accuracy after augmenting the training set with generated samples. The headline result is that window-size-3 augmentation improves classification accuracy from 57.6% to 87.6% (Table II), compared to 77.2% with traditional augmentation.

Significance. If the reported augmentation gain is robust, the proposed architecture could be practically useful for small sensor datasets, and its simplicity is a genuine strength. The paper is also commendably explicit about the limitations of its evaluation, acknowledging in the conclusion that no criterion is provided to quantitatively demonstrate the authenticity of the generated data. However, the empirical support for the central claim of "high-quality" generation is substantially weakened by a data-leakage error in the realism check, the absence of modern deep-generative baselines, and the lack of any fidelity metric on full sequences. The paper is a plausible proof-of-concept, but the evidence as presented is not sufficient to establish the stated claims.

major comments (5)
  1. [Section IV-E, Figures 8] The realism check for generated full sequences is contaminated: the text states that "the classification model, trained on both the training and testing datasets, was then used to classify the generated data." Because the testing set was used in training this classifier, the reported accuracies (84% and 94.8% for window sizes 3 and 5) cannot be interpreted as evidence that the generated sequences resemble real data. The classifier must be retrained using only the training split, and the resulting accuracies reported.
  2. [Sections II-B2 and III-B2] The entire generation pipeline rests on the premise that "the value at the first time step influences the readings at all subsequent time steps," but this premise is neither tested nor argued for the sign-language dataset. The dataset includes six dynamic signs with mid-gesture transitions and variability, which are not obviously determined by the initial frame. The paper provides no diagnostic comparing the temporal dynamics of generated and real sequences (e.g., autocorrelation, dynamic time warping, or subsequence-level statistics). Without such a check, the architecture's core assumption remains unsupported, and "high-quality augmented data" is not established.
  3. [Section IV-E, Tables II and III] The evaluation lacks statistical rigor and a proper model-selection protocol. All results are single-run with no standard deviations or repeated-seed reporting, and the interval weights, window size, and diffusion steps are tuned on the same dataset that is later used for the classification benchmark. It is therefore impossible to distinguish the claimed +30.0 improvement from run-to-run variability or selection overfitting. The paper should report error bars across multiple seeds and describe how hyperparameters were selected on a validation split.
  4. [Section IV-C, Table I] The FID metric is computed only on the 34-dimensional first time-step vectors, not on full sequences. The paper itself acknowledges that "FID metric cannot effectively capture the temporal dependencies of long time series" (Section IV). Since the central claim concerns the quality of entire augmented time series, FID on the first step does not provide support for that claim. A fidelity metric on full sequences, such as MMD on subsequences or a classifier trained only on the training set, is needed.
  5. [Section V (Conclusion)] The conclusion states that "it is crucial to find a criterion that helps the model quantitatively demonstrate the authenticity and effectiveness of the generated data," which concedes that the paper lacks such a criterion. This is a load-bearing limitation rather than a minor future-work item, because the abstract and introduction repeatedly claim "high-quality" generation. The paper should either provide a full-sequence fidelity evaluation or substantially soften the central claim.
minor comments (6)
  1. [Section III-B3] The interval weights w_1 to w_4 are introduced but their numerical values are never reported; please provide them and a sensitivity analysis.
  2. [Sections III-A and III-B1] The symbol T is used for both the number of diffusion steps (Section III-A) and the sequence length in the Transformer input (Section III-B1). Please disambiguate to avoid confusion.
  3. [Algorithm 1] The decoder input is set to the same tensor as the encoder input (X[:, 0:609, :]) while the label is X[:, 1:610, :]. Given the discussion of the causal mask, please clarify how teacher forcing and the causal mask are applied during training, and specify how generation is performed autoregressively at inference time.
  4. [Section II-B3 and elsewhere] There are several typos and grammatical errors, including "generate generate" (Section II-B3), "convoltional" (Section III-A), inconsistent spacing in "V AE", and "Each classes was split" in the Table I caption. A careful proofread is needed.
  5. [Section IV-E, Figure 8] The text refers to confusion matrices for each window size, but the subfigures in Figure 8 do not visibly contain confusion matrices. Please add the matrices or remove the reference.
  6. [Section II-B2] The citations [23] and [24] (Kingma and Box et al.) do not directly support the strong claim that the value at the first time step influences all subsequent time steps; this is presented as a general autoregressive property but is more specific. Please cite a source that explicitly supports this assumption or soften the wording.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the central claim is an empirical downstream-accuracy result on a held-out test set, not a derivation that reduces to its inputs.

full rationale

The paper does not attempt a first-principles derivation; its core claim is that diffusion-generated first time-steps plus transformer-generated continuations improve sign-language classification when added to the training set. That claim is checked against a separate test set: the first dataset trains the augmentation model and the classifier, and the second dataset is used only for final evaluation, so the reported +30.0 accuracy gain in Table II is a direct external measurement rather than a quantity recovered from a fitted parameter. The design premise that 'the value at the first time step influences the readings at all subsequent time steps' (Section II-B2) is explicitly presented as an assumption and is attributed to external autoregressive-modeling references, not to the authors' own prior work, so no self-citation chain is load-bearing. The weighted-loss intervals and window size are selected using the same data, which is a statistical-selection concern, but it does not make the held-out classification improvement equivalent to the model's inputs by construction. The contaminated realism check in Section IV-E, where 'the classification model, trained on both the training and testing datasets, was then used to classify the generated data,' weakens the 'closely resemble real data' interpretation and the conclusion's own caveat admits that 'it is crucial to find a criterion that helps the model quantitatively demonstrate the authenticity and effectiveness of the generated data.' These are validity limitations, not circular reductions: no equation in the paper is shown to equal its own input, no fitted parameter is renamed as a prediction, and no uniqueness or ansatz result is imported from the authors' prior publications. The evaluation logic is therefore self-contained rather than circular.

Assumptions & free parameters 6 free parameters · 4 assumptions · 1 invented entities

The central claim rests on a domain assumption about first-step determinism, an ad hoc loss-weighting procedure fitted to the collected data, and standard DDPM and Transformer machinery. All hyperparameters are dataset-specific, and no independent benchmark or external dataset validates the method's generality.

free parameters (6)
  • Interval weights w_1 to w_4 = not reported
    Assigned based on quartile boundaries of the training data's averaged difference curve; no values or general selection rule are given, so they are fitted to this dataset (Eq. 6).
  • Transformer window size = 3 and 5
    Hyperparameter controlling the view mask; results for 1, 3, and 5 are reported, and 'optimal parameters' are selected using the evaluation setup.
  • Diffusion time steps T = 10
    Chosen for the 34-dimensional data because the authors found more steps can reduce quality; not derived from theory.
  • Noise standard deviation for transformer input = 0.1
    Added to the first time steps for robustness, motivated by VAE, but not tuned or justified with an experiment (Eq. 8).
  • Polynomial degree for difference-curve fit = 20
    Used to fit the averaged inter-sample difference plot to compute quartiles for interval splitting; an arbitrary choice.
  • Training schedule = 200 epochs at 1e-4, 100 epochs MSE, 100 epochs at 1e-5
    Learning rates and epoch counts are chosen by hand; no ablation is provided.
assumptions (4)
  • domain assumption The value at the first time step determines the values at all subsequent time steps for the generated series.
    Explicitly stated in Section II-B2 as a premise of the design and used to justify generating only the initial time step with the diffusion model.
  • ad hoc to paper Padded no-change segments in sign-language actions can be detected from the averaged inter-sample difference curve and split into four intervals via quartiles.
    Section III-B3 uses a 20-degree polynomial fitted to the averaged difference plot and quartile boundaries to define loss weights; this is specific to the collected dataset's padding convention.
  • domain assumption A DDPM forward and reverse process with direct original-data prediction is a valid generative model for 34-dimensional initial time-step vectors.
    The authors modify Stable Diffusion by replacing convolutions with linear layers and predicting the data vector rather than noise (Section III-A); no formal guarantee or external validation is given.
  • domain assumption Class labels are available as conditioning information for both the diffusion and transformer components.
    The model uses label and position embeddings; this assumes labeled training data for the target domain, which is not always available in time-series augmentation settings.
invented entities (1)
  • View Mask matrix
    purpose: Restricts encoder self-attention to the previous k time steps (window size) when predicting each position.
    Introduced in Section III-B1 and Figure 5 as a new architectural component. No ablation is provided to isolate its contribution, and no external falsifiable prediction is made.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Time-Series Data Augmentation Model through Diffusion and Transformer Integration." pith.science (2026). https://pith.science/paper/X74C2M7W

@misc{pith2026250503790,
  author       = {Pith},
  title        = {Pith review of: A Time-Series Data Augmentation Model through Diffusion and Transformer Integration},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/X74C2M7W}},
  note         = {Machine review of arXiv:2505.03790}
}
read the original abstract

With the development of Artificial Intelligence, numerous real-world tasks have been accomplished using technology integrated with deep learning. To achieve optimal performance, deep neural networks typically require large volumes of data for training. Although advances in data augmentation have facilitated the acquisition of vast datasets, most of this data is concentrated in domains like images and speech. However, there has been relatively less focus on augmenting time-series data. To address this gap and generate a substantial amount of time-series data, we propose a simple and effective method that combines the Diffusion and Transformer models. By utilizing an adjusted diffusion denoising model to generate a large volume of initial time-step action data, followed by employing a Transformer model to predict subsequent actions, and incorporating a weighted loss function to achieve convergence, the method demonstrates its effectiveness. Using the performance improvement of the model after applying augmented data as a benchmark, and comparing the results with those obtained without data augmentation or using traditional data augmentation methods, this approach shows its capability to produce high-quality augmented data.

Figures

Figures reproduced from arXiv: 2505.03790 by the authors.

Figure 1
Figure 1. Analysis domain in signal processing data generation tasks by adjusting model configurations and hyperparameters. To evaluate the performance of our model, we designed a data glove equipped with pressure sensors to collect sign language data and applied data augmentation to the collected sign language data. In the experiments, we found that due to the characteristic of the MSE loss function, which tends to minimize … view at source ↗
Figure 2
Figure 2. Flowchart of Generative Model new data [22]. Notable examples include Gaussian Mixture Models(GMM), Hidden Markov Models(HMM), and Autore￾gressive Models(AR). However, the performance of data gen￾erated by explicit probability models tends to be relatively poor because most real-world data do not adhere to a fixed structure and cannot be modeled by a specific explicit probability function. Using simple assumptions m… view at source ↗
Figure 3
Figure 3. Diffusion Structure step, and d indicates the dimension of the feature vector. The Diffusion model is used to generate the initial time-step data for an arbitrary number N. These data are then fed into the Transformer model, which iteratively predicts the data for the next time step until the desired input sequence length is achieved. A. Diffusion Model Because we followed the baseline theory that once the data at t… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Transformer Structure B. Transformer Model The Transformer is used to predict the data after the first time step. Although termed ”prediction”, it actually involves generating the most likely data readings based on the distri￾bution of the existing real data, given the…
Figure 5
Figure 5. Figure 5: View Mask View Mask: At the end of encoder, we add a view mask. The view mask ∈ R T ×T . It is a two-dimensional matrix, with both the horizontal and vertical axes representing the sequence length. The view mask is used to mask out data from positions that do not need …
Figure 6
Figure 6. Figure 6: The division of intervals. the finger movements [28], [29]. The location of key point is shown in Figure 7b. And The sewn glove is shown in Figure 7c. In addition to the pressure sensors, there is an external controller that processes the sensor data and transmits it t…
Figure 7
Figure 7. Figure 7: Design and Fabrication of the Data Glove [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Classification Results of Generated Data Across Dif [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 12 canonical work pages

  1. [1]

    Small data challenges of studying rare diseases,

    A. A. Mitani and S. Haneuse, “Small data challenges of studying rare diseases,” JAMA network open , vol. 3, no. 3, pp. e201 965–e201 965, 2020

    work page 2020

  2. [2]

    Clinical research for rare disease: opportunities, challenges, and solutions,

    R. C. Griggs, M. Batshaw, M. Dunkle, R. Gopal-Srivastava, E. Kaye, J. Krischer, T. Nguyen, K. Paulus, P. A. Merkel et al., “Clinical research for rare disease: opportunities, challenges, and solutions,” Molecular genetics and metabolism , vol. 96, no. 1, pp. 20–26, 2009

    work page 2009

  3. [3]

    Skeleton-based data augmentation for sign language recognition using adversarial learning,

    Y . Nakamura and L. Jing, “Skeleton-based data augmentation for sign language recognition using adversarial learning,” IEEE Access, 2024

    work page 2024

  4. [4]

    The variational fair autoencoder,

    C. Louizos, K. Swersky, Y . Li, M. Welling, and R. Zemel, “The variational fair autoencoder,” arXiv preprint arXiv:1511.00830 , 2015

    arXiv 2015

  5. [5]

    Generative adversarial nets,

    I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y . Bengio, “Generative adversarial nets,” Advances in neural information processing systems , vol. 27, 2014

    2014

  6. [6]

    Informer: Beyond efficient transformer for long sequence time-series forecasting,

    H. Zhou, S. Zhang, J. Peng, S. Zhang, J. Li, H. Xiong, and W. Zhang, “Informer: Beyond efficient transformer for long sequence time-series forecasting,” in Proceedings of the AAAI conference on artificial intel- ligence, vol. 35, no. 12, 2021, pp. 11 106–11 115

    2021

  7. [7]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in neural information processing systems , vol. 33, pp. 6840– 6851, 2020

    2020

  8. [8]

    Deep unsupervised learning using nonequilibrium thermodynamics,

    J. Sohl-Dickstein, E. Weiss, N. Maheswaranathan, and S. Ganguli, “Deep unsupervised learning using nonequilibrium thermodynamics,” in International conference on machine learning . PMLR, 2015, pp. 2256–2265

    2015

Show all 29 references
  1. [9]

    Zero-shot text-to-image generation,

    A. Ramesh, M. Pavlov, G. Goh, S. Gray, C. V oss, A. Radford, M. Chen, and I. Sutskever, “Zero-shot text-to-image generation,” in International conference on machine learning . Pmlr, 2021, pp. 8821–8831

    2021

  2. [10]

    High- resolution image synthesis with latent diffusion models,

    R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High- resolution image synthesis with latent diffusion models,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2022, pp. 10 684–10 695

    2022

  3. [11]

    Diffusion-lm improves controllable text generation,

    X. Li, J. Thickstun, I. Gulrajani, P. S. Liang, and T. B. Hashimoto, “Diffusion-lm improves controllable text generation,” Advances in Neu- ral Information Processing Systems , vol. 35, pp. 4328–4343, 2022. IEEE TRANSACTIONS ON HUMAN-MACHINE SYSTEMS 10

    2022

  4. [12]

    Dif- fwave: A versatile diffusion model for audio synthesis,

    Z. Kong, W. Ping, J. Huang, K. Zhao, and B. Catanzaro, “Dif- fwave: A versatile diffusion model for audio synthesis,” arXiv preprint arXiv:2009.09761, 2020

    2009 arXiv

  5. [13]

    Real-valued (medical) time series generation with recurrent conditional gans,

    C. Esteban, S. L. Hyland, and G. R ¨atsch, “Real-valued (medical) time series generation with recurrent conditional gans,” arXiv preprint arXiv:1706.02633, 2017

    2017 arXiv

  6. [14]

    Diffusion-ts: Interpretable diffusion for general time series generation,

    X. Yuan and Y . Qiao, “Diffusion-ts: Interpretable diffusion for general time series generation,” arXiv preprint arXiv:2403.01742 , 2024

    2024 arXiv

  7. [15]

    Stl: A seasonal-trend decomposition,

    R. B. Cleveland, W. S. Cleveland, J. E. McRae, I. Terpenning et al. , “Stl: A seasonal-trend decomposition,” J. off. Stat , vol. 6, no. 1, pp. 3–73, 1990

    1990

  8. [16]

    Data augmentation for time series classification using convolutional neural networks,

    A. Le Guennec, S. Malinowski, and R. Tavenard, “Data augmentation for time series classification using convolutional neural networks,” in ECML/PKDD workshop on advanced analytics and learning on temporal data, 2016

    2016

  9. [17]

    A review: Data pre- processing and data augmentation techniques,

    K. Maharana, S. Mondal, and B. Nemade, “A review: Data pre- processing and data augmentation techniques,” Global Transitions Pro- ceedings, vol. 3, no. 1, pp. 91–99, 2022

    2022

  10. [18]

    Time series data augmentation for deep learning: A survey,

    Q. Wen, L. Sun, F. Yang, X. Song, J. Gao, X. Wang, and H. Xu, “Time series data augmentation for deep learning: A survey,” arXiv preprint arXiv:2002.12478, 2020

    2002 arXiv

  11. [19]

    Wavelet transform application for/in non-stationary time-series analysis: A re- view,

    M. Rhif, A. Ben Abbes, I. R. Farah, B. Mart ´ınez, and Y . Sang, “Wavelet transform application for/in non-stationary time-series analysis: A re- view,” Applied Sciences, vol. 9, no. 7, p. 1345, 2019

    2019

  12. [20]

    What is the fast fourier transform?

    W. T. Cochran, J. W. Cooley, D. L. Favin, H. D. Helms, R. A. Kaenel, W. W. Lang, G. C. Maling, D. E. Nelson, C. M. Rader, and P. D. Welch, “What is the fast fourier transform?” Proceedings of the IEEE , vol. 55, no. 10, pp. 1664–1674, 1967

    1967

  13. [21]

    Detecting trend and seasonal changes in satellite image time series,

    J. Verbesselt, R. Hyndman, G. Newnham, and D. Culvenor, “Detecting trend and seasonal changes in satellite image time series,” Remote sensing of Environment, vol. 114, no. 1, pp. 106–115, 2010

    2010

  14. [22]

    Learning deep generative models,

    R. Salakhutdinov, “Learning deep generative models,” Annual Review of Statistics and Its Application , vol. 2, no. 1, pp. 361–385, 2015

    2015

  15. [23]

    Auto-encoding variational bayes,

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

    2013 arXiv

  16. [24]

    G. E. Box, G. M. Jenkins, G. C. Reinsel, and G. M. Ljung, Time series analysis: forecasting and control . John Wiley & Sons, 2015

    2015

  17. [25]

    Dataset augmentation in feature space,

    T. DeVries and G. W. Taylor, “Dataset augmentation in feature space,” arXiv preprint arXiv:1702.05538 , 2017

    2017 arXiv

  18. [26]

    Reinforcement learning: A survey,

    L. P. Kaelbling, M. L. Littman, and A. W. Moore, “Reinforcement learning: A survey,” Journal of artificial intelligence research , vol. 4, pp. 237–285, 1996

    1996

  19. [27]

    A survey of transfer learning,

    K. Weiss, T. M. Khoshgoftaar, and D. Wang, “A survey of transfer learning,” Journal of Big data , vol. 3, pp. 1–40, 2016

    2016

  20. [28]

    Sign language vocabulary recognition only with tactile sensing glove,

    M. Kagami, Z. Yu, S. T. Ceng, and L. Jing, “Sign language vocabulary recognition only with tactile sensing glove,” in 2024 IEEE 17th Interna- tional Symposium on Embedded Multicore/Many-core Systems-on-Chip (MCSoC). IEEE, 2024, pp. 110–116

    2024

  21. [29]

    Capturing com- plex hand movements and object interactions using machine learning- powered stretchable smart textile gloves,

    A. Tashakori, Z. Jiang, A. Servati, S. Soltanian, H. Narayana, K. Le, C. Nakayama, C.-l. Yang, Z. J. Wang, J. J. Eng et al., “Capturing com- plex hand movements and object interactions using machine learning- powered stretchable smart textile gloves,” Nature Machine Intelligen...

    2024

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.