Pith. sign in

REVIEW 4 major objections 6 minor 44 references

Hybrid Adversarial Spectral Loss Conditional Generative Adversarial Networks for Signal Data Augmentation in Ultra-precision Machining Surface Roughness Prediction

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · 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 augmenting a 52-sample ultra-precision machining dataset with 520 generated force signals lowers surface roughness prediction error from 31.4% to about 8.8% MAPE.

desk verdict Useful empirical comparison of CGAN variants for 1D force signals, but the headline augmentation claim rests on an unspecified signal-to-parameter pairing and a single small split. read the letter →

arxiv 2507.04665 v1 pith:SLQFHZ2V submitted 2025-07-07 cs.LG

Suiyan Shang , Chi Fai Cheung , Pai Zheng This is my paper · ORCID
classification cs.LG
keywords signaldataaugmentationconditionalgenerativeadversarialnetworkspectrallosssurfaceroughnesspredictionultra-precisionmachiningwaveletcoherenceCNN-Transformer
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to establish that a conditional generative adversarial network with an added frequency-domain spectral loss can generate realistic ultra-precision machining force signals, and that those synthetic signals can serve as effective extra training data for surface roughness prediction. On a 64-sample milling dataset, the authors compare five CGAN variants and report that their hybrid adversarial spectral loss reproduces high-frequency content best, with wavelet coherence above 0.85. Adding ten times as many generated samples lowers the CNN-Transformer mean absolute percentage error from 31.4% to about 8.8%, with diminishing returns beyond that. This matters because real ultra-precision machining data is scarce and expensive to collect; if synthetic force signals carry usable physical information, accurate quality-control models could be trained without extensive additional machining experiments.

What carries the argument

The central object is the hybrid adversarial spectral loss for the generator, defined as $HAS\_Loss_G = \gamma_1 Loss_1 + \gamma_2 Loss_2$ with $\gamma_1 + \gamma_2 = 1$, where $Loss_1$ is the standard conditional GAN generator loss and $Loss_2$ averages, over the batch and time frames, the squared Frobenius norm of the difference between the STFT magnitudes of real and generated force signals. This spectral term constrains the generator in the frequency domain so that high-frequency content is not lost. The generator itself is a three-layer 1D convolution-transpose network that takes sinusoidal noise combined with a surface-roughness label as input, while the discriminator is a three-layer 1D convolutional network that judges whether a signal is real or generated under the same label condition.

What would settle it

A concrete check would be to train the same CNN-Transformer predictor after randomly permuting the machining parameters attached to the generated signals while keeping the roughness labels fixed; if the MAPE stays near 8.8%, the improvement does not depend on an intact signal-parameter relationship and the central claim is weakened.

Watch

Extended reading notes

Core claim

The paper's central claim is that a lightweight 1D convolutional conditional GAN, trained with a generator loss that penalizes the difference between the short-time Fourier transform magnitudes of real and generated signals, produces synthetic force signals that are more faithful to real machining signals than those from plain CGAN, convolutional CGAN, ACGAN, or WCGAN. The improvement is most visible for high-frequency signals, where the spectral loss directly punishes fitting error in the Fourier domain. When these generated signals are combined with machining parameters and added to the training set, end-to-end predictors that extract features automatically improve substantially, with the best model reaching about 8.8% MAPE compared with 31.4% on the original 52 real samples. The paper concludes that for 1D industrial signals, simpler GAN architectures with frequency-aware losses are more suitable than theoretically more complex variants, and that CGAN-based augmentation is a viable route toward real-time virtual metrology for ultra-precision machining.

Load-bearing premise

The load-bearing premise is that each synthetic force signal can be paired with machining parameters such as spindle speed, feed rate, and depth of cut in a way that preserves the true signal-parameter-roughness relationship, but the paper never states where those parameter values come from or how the pairing is made.

Editorial extensions

If this is right

  • End-to-end models that learn features from raw waveforms benefit from generated signals, while models that rely on hand-crafted time- and frequency-domain features do not; augmentation helps only when the predictor can use the raw signal structure.
  • A roughly tenfold augmentation, or about 520 generated samples, is the practical ceiling for this dataset; beyond it the prediction error plateaus near 9% MAPE, so extra generation yields diminishing returns.
  • The spectral loss is what improves high-frequency fidelity; the paper reports that CGAN variants without it fail to reproduce the amplitudes of high-frequency force signals.
  • If the result holds, CGAN-based augmentation offers a route toward virtual metrology for ultra-precision machining, reducing dependence on time-consuming offline surface measurements.

Reading between the lines

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

  • Editorial extension: because the generator conditions only on sinusoidal noise and a roughness label, the paper leaves open how each synthetic signal is paired with spindle speed, feed rate, and depth of cut; conditioning the generator on those parameters explicitly would make the augmentation mechanism more transparent and testable.
  • Editorial extension: part of the reported improvement could come from the predictor seeing more examples spread across the label range rather than from physically faithful waveforms; a control experiment using randomly relabeled or noise-only augmented data would separate these effects.
  • Editorial extension: the same spectral-loss recipe may transfer to other small-dataset industrial signal problems, but the optimal augmentation ratio is likely to depend on signal dimensionality and label diversity rather than being a universal tenfold rule.
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

4 major / 6 minor

Summary. The paper proposes HAS-CGAN, a conditional GAN augmented with a spectral loss, to generate synthetic 1D cutting-force signals for data augmentation in ultra-precision machining surface roughness prediction. The authors compare five CGAN variants on signal fidelity using wavelet coherence, then combine generated signals with machining parameters to enlarge a 52-sample training set and report that training CNN-Transformer on a 10x-augmented dataset reduces MAPE from 31.4% to 8.8%. The central claim is that the generated signals are realistic enough to serve as effective training data.

Significance. If substantiated, the approach would be practically valuable for small-sample industrial datasets, where collecting and labeling sensor data is expensive. The paper has some strengths: it systematically compares five GAN variants, uses a quantitative fidelity metric (wavelet coherence), and evaluates augmentation across multiple prediction architectures. However, the central claim is not supported as written. The generator is not conditioned on machining parameters, yet the augmented dataset pairs each synthetic signal with machining parameters; this gap undermines the validity of the reported MAPE improvement. Internal inconsistencies in dataset counts and augmentation sizes, the circularity of the wavelet-coherence fidelity claim, and the absence of any uncertainty quantification further reduce confidence. The manuscript is not ready for publication in its current form.

major comments (4)
  1. [§2.2.1 (Eqs. 1–6) and §3.3] The generator takes only sinusoidal noise and a surface-roughness label as input; spindle speed, feed rate, and depth of cut never enter the generation process. Yet §2.1 states that generated signals are provided "with corresponding machining parameters," and §3.3 trains predictors on (signal, parameters, Ra) triples. The paper never specifies how machining parameters are assigned to synthetic signals. If the assignment is arbitrary or based only on the Ra label, the augmented triples are not physically coherent because the generated waveform has no parameter dependence. The reported MAPE improvement from 31.4% to 8.8% could then reflect label/artifact correlations rather than a learned signal-parameter-roughness mapping. This is load-bearing because the central claim that data augmentation improves prediction is meaningless without valid triples.
  2. [§3.1 and §3.3] The training-set size is stated as 52 samples in §3.1, but §3.3 refers to "the original 56 real samples." Similarly, the 5-times augmentation is described as "210 samples" in the text but as "260 samples" in the Fig. 3.c caption. Since 10x = 520 is consistent with 52 samples, the 5x and 56-sample counts are internally contradictory. These inconsistencies make the exact experimental protocol ambiguous and undermine the quantitative precision of the headline result.
  3. [§3.2 and Eq. (5)] Wavelet coherence is used as the principal fidelity metric, with conflicting thresholds reported in the abstract (>0.85), conclusion (>0.9), and text (>0.8). More importantly, the high WC for HAS-CGAN is partly by construction: the generator's custom loss in Eq. (5) directly minimizes the magnitude difference between the STFT of real and generated signals, and wavelet coherence measures time-frequency similarity. Without a baseline comparison (e.g., WC between two real signals, or WC for a non-spectral-loss CGAN), the fidelity gain could be an artifact of explicit frequency-domain matching rather than genuine signal realism. The downstream MAPE reduction is an independent outcome, but the generation-fidelity claim is not established.
  4. [§3.3] The evaluation relies on a single random 52/12 split with no confidence intervals, no repeated runs, and no code. The improvement in MAPE from 31.4% to 8.8% is based on a single 12-sample test set; the paper does not report variance across seeds or any statistical significance test. Without this, the improvement is anecdotal rather than established, especially given the very small test set.
minor comments (6)
  1. [§3.1] The paper refers to "HAS-CGAN (Hierarchical Attention-Supervised Conditional Generative Adversarial Network)", which contradicts the title's "Hybrid Adversarial Spectral Loss"; please use one consistent name.
  2. [Eq. (5)] The STFT subscripts/superscripts for the real and generated signals are difficult to distinguish; both appear as x_i with different labels, and the notation should be clarified.
  3. [§2.1 vs §3.2] The generator is described as using "three 1D-convolutional transpose layers" in §2.1 but as a "3-layer fully connected network" in §3.2; please reconcile this discrepancy.
  4. [§3.3] The order of methods is inconsistent: "SVR, LSTM and RF" appears in one paragraph and "SVR, RF and LSTM" in the next; keep the ordering consistent throughout.
  5. [Abstract/Conclusions] The wavelet-coherence claims differ across the abstract (>0.85), conclusion (>0.9), and body (>0.8); report the actual values or a range associated with the relevant figure.
  6. [General] There are numerous typographical and grammatical errors (e.g., "no matther," "augemented," "resepectively") that interfere with readability; a thorough language edit is needed.

Circularity Check

1 steps flagged · score 4.0 of 10

Wavelet-coherence fidelity is partly enforced by the spectral loss; the downstream MAPE gain is independent but depends on an unspecified parameter pairing.

  1. fitted input called prediction [Section 2.2.1, Eq. (5); Section 3.2, Fig. 2; Abstract]
    "L_Loss2 = (1/M)(1/T) Σ || |STFT(x_real)| − |STFT(x_generated)| ||_F^2 ... when adding the spectral loss, which is computed in the frequency domain ... our proposed method with spectral loss punishment increases the coherence of generated signals and original signals to around 0.9. Abstract: 'achieving >0.85 wavelet coherence through Fourier-domain optimization.'"

    The generator's custom loss directly penalizes the difference between the STFT magnitudes of generated and real signals, i.e. a time-frequency-domain fit on the real training signals. The paper then reports wavelet coherence (WC), also a time-frequency similarity metric, as evidence that HAS-CGAN 'excels' in high-frequency generation. Because the loss was explicitly designed to minimize frequency-domain error on these same signals, the high WC is a restatement of the optimized objective on the fitting set, not an independent prediction.

full rationale

The paper's central quantitative claim — that adding 520 generated force signals lowers CNN-Transformer MAPE from 31.4% to 8.8% — is an independent, externally evaluated outcome: the predictor is trained on augmented triples and tested on 12 held-out real samples. That part does not reduce to the generator loss or to a fit, so the core augmentation result is not circular. The one significant circular-adjacent step is the fidelity headline: Eq.5 trains the generator by minimizing STFT magnitude error against real signals, and the paper then reports high wavelet coherence on those same signals as evidence that HAS-CGAN 'excels' at high-frequency generation. Since both the loss and the evaluation metric are time-frequency similarity measures computed on the fitting data, the >0.85 WC result is substantially a restatement of the optimization objective rather than an independent validation. This is a fitted-input-called-prediction step. Separately, Section 2.1 says augmented data contains generated signals 'with corresponding machining parameters,' but the generator (Eqs.1-3) conditions only on sinusoidal noise and an Ra label; how spindle speed, feed rate, and depth of cut are paired with synthetic signals is never specified. This is a serious correctness gap in the augmented triple construction, though not itself a circular reduction. Self-citations (refs. 14, 41) provide the dataset and preprocessing but are not load-bearing theoretical premises.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The method rests on several unstated domain assumptions: that generated signals can be paired with machining parameters they were not conditioned on, that a 12-sample test set is reliable, and that STFT/wavelet measures capture what matters for roughness prediction. The only explicit free parameters are the hybrid loss weights, which are not reported.

free parameters (3)
  • gamma_1 / gamma_2 (hybrid loss weights) = not reported
    Eq. 6 defines generator loss as gamma1*L_GAN1 + gamma2*L_spectral with gamma1+gamma2=1. The split is never given, yet it controls the balance between adversarial and spectral fidelity and is central to the method.
  • Augmentation scale (number of generated samples) = 520 (10x)
    The paper scans 5x, 7x, 10x, 15x, 20x and selects 10x as optimal based on test-set MAPE plateau, so the headline 8.8% is a selected result rather than an a priori setting.
  • Kernel size and filter widths = kernel 20, filters 16/32/64
    Architecture hyperparameters are chosen by hand to match 'approximate periodicity' and are not ablated; the central comparison between GAN variants depends on these choices.
assumptions (5)
  • domain assumption Force signals from UPM are periodic and decomposable into sinusoids, justifying sinusoidal noise.
    Section 2.1 states sinusoidal noise is theoretically justified by the periodic nature of force signals; if false, the generator's input distribution is mismatched.
  • domain assumption Generated signals can be paired with machining parameters that were not part of the generator input without breaking the signal-parameter-roughness relationship.
    Section 3.3 builds augmented data from 'generated signals with corresponding machining parameters and labels', but the generator in Section 2.2.1 sees only noise and label. The source of the parameters is unspecified and load-bearing.
  • domain assumption Wavelet coherence is a valid measure of generation fidelity for 1D force signals.
    Section 3.2 uses WC as the quantitative similarity metric; no validation that WC correlates with downstream prediction usefulness.
  • domain assumption The 52/12 random split of 64 samples is representative enough for stable MAPE estimates.
    Section 3.1 defines the split; with 12 test samples, MAPE has large variance and no error bars are reported.
  • domain assumption STFT magnitude difference is a suitable training objective for statistically faithful signals.
    Eq. 5 defines the spectral loss; standard in audio but assumed here as sufficient for force signals.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Hybrid Adversarial Spectral Loss Conditional Generative Adversarial Networks for Signal Data Augmentation in Ultra-precision Machining Surface Roughness Prediction." pith.science (2026). https://pith.science/paper/SLQFHZ2V

@misc{pith2026250704665,
  author       = {Pith},
  title        = {Pith review of: Hybrid Adversarial Spectral Loss Conditional Generative Adversarial Networks for Signal Data Augmentation in Ultra-precision Machining Surface Roughness Prediction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SLQFHZ2V}},
  note         = {Machine review of arXiv:2507.04665}
}
read the original abstract

Accurate surface roughness prediction in ultra-precision machining (UPM) is critical for real-time quality control, but small datasets hinder model performance. We propose HAS-CGAN, a Hybrid Adversarial Spectral Loss CGAN, for effective UPM data augmentation. Among five CGAN variants tested, HAS-CGAN excels in 1D force signal generation, particularly for high-frequency signals, achieving >0.85 wavelet coherence through Fourier-domain optimization. By combining generated signals with machining parameters, prediction accuracy significantly improves. Experiments with traditional ML (SVR, RF, LSTM) and deep learning models (BPNN, 1DCNN, CNN-Transformer) demonstrate that augmenting training data with 520+ synthetic samples reduces prediction error from 31.4% (original 52 samples) to ~9%, effectively addressing data scarcity in UPM roughness prediction."

Figures

Figures reproduced from arXiv: 2507.04665 by the authors.

Figure 1
Figure 1. Schematic illustration of Conditional Generative Adversarial Networks for signal data augmentation in ultra-precision surface roughness prediction. a. The overall architecture of our proposed HAS-CGAN for signal data augmentation in ultra-precision surface roughness prediction. b. The flowchart of the HAS-GAN training process for each epoch. error penalization. Our experimental results conclusively demonstrate that … view at source ↗
Figure 1
Figure 1. Fig.1.b [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. The illustration of visual comparison and quantitative similarity computation of generated signals and real signals. a. The time domain waveform of force signals from generated samples and real samples under various conditional CGANs. b. The wavelet coherence of generated force signals and real force signals. similarity to real experimental data. Besides the HAS￾CGAN, Auxiliary Conditional GAN can also use an auxili… view at source ↗
Figures from the paper (3 more)
Figure 3
Figure 3. Figure 3: The schematic illustration of the comparison of prediction results before and after data augmentation under various prediction methods and different times of augmentation datasets. a. Prediction performance under different times of training data from various prediction…
Figure 2
Figure 2. Figure 2: b [PITH_FULL_IMAGE:figures/full_fig_p007_2.png]
Figure 3
Figure 3. Figure 3: c [PITH_FULL_IMAGE:figures/full_fig_p008_3.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

44 extracted references · 40 canonical work pages

  1. [1]

    Asiltürk, I., & Akkuş, H. (2011). Determining the effect of cutting parameters on surface roughness in hard turning using the Taguchi method. Measurement, 44(9), 1697-1704

    work page 2011

  2. [2]

    B., & Cheung, B

    Lee, W. B., & Cheung, B. C. (2003). Surface generation in ultra-precision diamond turning: modelling and practices (Vol 12). John Wiley & Sons

    work page 2003

  3. [3]

    N., & Bissacco, G

    Alting, L., Kimura, F., Hansen, H. N., & Bissacco, G. (2003). Micro engineering. CIRP annals, 52(2), 635-657

    work page 2003

  4. [4]

    Kumar, P. (2015). Design and Development of an Optical Chip Interferometer for High Precision On -Line Surface Measurement (Doctoral dissertation, University of Huddersfield)

    work page 2015

  5. [5]

    Shieh, H. P. D., Huang, Y. P., & Chien, K. W. (2005). Micro-optics for liquid crystal displays applications. Journal of display technology, 1(1), 62

    work page 2005

  6. [6]

    Gao, W., Haitjema, H.H., Fang, F., Leach, R.K., Cheung, C.F., Savio, E., & Linares, J. (2019). On -machine and in - process surface metrology for precision manufacturing. CIRP Annals

    work page 2019

  7. [7]

    Li, Z., Dai, Y., Guan, C., Lai, T., Sun, Z., & Li, H. (2024). An on -machine measurement technique with sub -micron accuracy on a low-precision grinding machine tool. Journal of Manufacturing Processes

    work page 2024

  8. [8]

    F.-T. Cheng. ( 2004). Researching Strategy and Development Proposal of e -Manufacturing. Automation Division of National Science Council. Taiwan, R.O.C

    work page 2004

Show all 44 references
  1. [9]

    Lin, T., Hung, M., Lin, R., & Cheng, F. (2006). A virtual metrology scheme for predicting CVD thickness in semiconductor manufacturing. Proceedings 2006 IEEE International Conference on Robotics and Automation, 2006. ICRA 2006., 1054-1059

    2006

  2. [10]

    Tieng, H., Li, Y., Tseng, K., Yang, H., & Cheng, F. (2020). An Automated Dynamic -Balancing-Inspection Scheme for Wheel Machining. IEEE Robotics and Automation Letters, 5, 2224-2231

    2020

  3. [11]

    Hsieh, Y., Lin, C., Yang, Y., Hung, M., & Cheng, F. (2019). Automatic Virtual Metrology for Carbon Fiber Manufacturing. IEEE Robotics and Automation Letters, 4 , 2730-2737

    2019

  4. [12]

    Ruan, P., Saxena, D., Cao, J., Liu, X., Wang, R., & Cheung, C. F. (2025). NASPrecision: Neural Architecture Search-Driven Multi-Stage Learning for surface roughness prediction in ultra-precision machining. Expert Systems with Applications, 262, 125540

    2025

  5. [13]

    Rao, P. K. (2013). Sensor-based monitoring and inspection of surface morphology in ultraprecision manufacturing processes. Oklahoma State University

    2013

  6. [14]

    F., & Zheng, P

    Shang, S., Wang, C., Liang, X., Cheung, C. F., & Zheng, P. (2023). Surface roughness prediction in ultra -precision milling: An extreme learning machine method with data fusion. Micromachines, 14(11), 2016

    2023

  7. [15]

    T., & Komanduri, R

    Cheng, C., Wang, Z., Hung, W., Bukkapatnam, S. T., & Komanduri, R. (2015). Ultra -precision machining process dynamics and surface quality monitoring. Procedia Manufacturing, 1, 607-618

    2015

  8. [16]

    Mao, C., Tang, X., Zou, H., Zhou, Z., & Yin, W. (2012). Experimental investigation of surface quality for minimum quantity oil –water lubrication grinding. The International Journal of Advanced Manufacturing Technology, 59, 93-100

    2012

  9. [17]

    ] Bai, L., Yang, Q., Cheng, X.et al. (2023). A hybrid physics data driven surface roughness prediction model for ultra-precision machining. Science China Technological Sciences. 66, 1289–1303

    2023

  10. [18]

    & Wang, C

    Wang, Y., He, Z., Xie, S., Wang, R., Zhang, Z., Liu, S., ... & Wang, C. (2024). Explainable prediction of surface roughness in multi -jet polishing based on ensemble regression and differential evolution method. Expert Systems with Applications, 249, 123578

    2024

  11. [19]

    ( 2024) A Novel Approach to Surface Roughness Virtual Sample Generation to Address the Small Sample Size Problem in Ultra-Precision Machining

    Liu R, Tian W. ( 2024) A Novel Approach to Surface Roughness Virtual Sample Generation to Address the Small Sample Size Problem in Ultra-Precision Machining. Sensors.; 24(11):3621

    2024

  12. [20]

    Wei, S., Zou, S., & Liao, F. (2020). A comparison on data augmentation methods based on deep learning for audio classification. In Journal of physics: Conference series (Vol. 1453, No. 1, p. 012085). IOP Publishing

    2020

  13. [21]

    Salamon J, Bello JP. (2017). Deep convolutional neural networks and data augmentation for environmental sound classification. IEEE Sig Process Lett; 24(3):279–83

    2017

  14. [22]

    Nanni L, Maguolo G, Paci M. (2020 ). Data augmentation approaches for improving animal audio classification. Ecol inform; 57:

    2020

  15. [23]

    Cui X, Goel V, Kingsbury B. (2015). Data augmentation for deep neural network acoustic modeling. IEEE/ACM Trans Audio Speech Lang Process; 23 (9):1469–77

    2015

  16. [24]

    W. -N. Hsu, Y. Zhang and J. Glass. (2017) . Unsupervised domain adaptation for robust speech recognition via variational autoencoder -based data augmentation. IEEE Automatic Speech Recognition and Understanding Workshop (ASRU) , Okinawa, Japan, 2017, pp. 16-23

    2017

  17. [25]

    Salakhutdinov, R., & Hinton, G. E. (2012). Deep Boltzmann Machines for Data Augmentation in Speech Recognition. Journal of Machine Learning Research, *13*(1), 101-120

    2012

  18. [26]

    & Bengio, Y

    Goodfellow, I., Pouget -Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., ... & Bengio, Y. (2014). Generative adversarial nets. Advances in neural information processing systems, 27

    2014

  19. [27]

    S. Liu, S. Li and H. Cheng. (2022) Towards an End-to - End Visual-to-Raw-Audio Generation With GAN. in IEEE Transactions on Circuits and Systems for Video Technology, vol. 32, no. 3, pp. 1299-1312, March

    2022

  20. [28]

    P., & Welling, M

    Kingma, D. P., & Welling, M. (2014). Auto -encoding variational Bayes. International Conference on Learning Representations (ICLR)

    2014

  21. [29]

    Salakhutdinov, R. (2015). Learning deep generative models. Annual Review of Statistics and Its Application , 2, 361-385

    2015

  22. [30]

    Lucic, M., Kurach, K., Michalski, M., Gelly, S., & Bousquet, O. (2018). Are GANs created equal? A large-scale study. Advances in Neural Information Processing Systems (NeurIPS), *31*, 698-707

    2018

  23. [31]

    Brock, A., Donahue, J., & Simonyan, K. (2019). Large scale GAN training for high fidelity natural image synthesis. International Conference on Learning Representations (ICLR)

    2019

  24. [32]

    Zhao, S., Song, J., & Ermon, S. (2017). InfoVAE: Information maximizing variational autoencoders. arXiv:1706.02262

    2017 arXiv

  25. [33]

    Towards Principled Methods for Training GANs

    Arjovsky & Bottou (2017). "Towards Principled Methods for Training GANs". ICLR

    2017

  26. [34]

    Convergence and Stability of GANs

    Kodali et al. (2017). "Convergence and Stability of GANs". arXiv:1705.08584

    2017 arXiv

  27. [35]

    Odena, A., Olah, C., & Shlens, J. (2017). Conditional image synthesis with auxiliary classifier GANs. ICML, 2642- 2651

    2017

  28. [36]

    Gulrajani, I., Ahmed, F., Arjovsky, M., Dumoulin, V., & Courville, A. (2017). Improved training of Wasserstein GANs. NeurIPS, 5767-5777

    2017

  29. [37]

    Arjovsky, M., Chintala, S., & Bottou, L. (2017). Wasserstein GAN. ICML, 214-223

    2017

  30. [38]

    Shao, Siyu & Wang, Pu & Yan, Ruqiang. (2019). Generative adversarial networks for data augmentation in machine fault diagnosis. Computers in Industry. 106. 85-93

    2019

  31. [39]

    Li, X., Metsis, V., Wang, H., & Ngu, A.H. (2022). TTS- GAN: A Transformer -based Time-Series Generative Adversarial Network. Conference on Artificial Intelligence in Medicine in Europe

    2022

  32. [40]

    Radpour, D., & Bheda, V. (2017). Conditional Generative Adversarial Networks for Emoji Synthesis with Word Embedding Manipulation. ArXiv, abs/1712.04421

    2017 arXiv

  33. [41]

    (2024) Roughness prediction of end milling surface for behavior mapping of digital twined machine tools [version 2; peer review: 2 approved, 1 approved with reservations]

    Shang S, Jiang G, Sun Z et al. (2024) Roughness prediction of end milling surface for behavior mapping of digital twined machine tools [version 2; peer review: 2 approved, 1 approved with reservations]. Digital Twin 2024, 3:4

    2024

  34. [42]

    Mirza, M., & Osindero, S. (2014). Conditional generative adversarial nets. arXiv preprint arXiv:1411.1784

    2014 arXiv

  35. [43]

    2016.DOI:10.48550/arXiv.1610.09585

    Odena A , Olah C , Shlens J .Conditional Image Synthesis With Auxiliary Classifier GANs[J]. 2016.DOI:10.48550/arXiv.1610.09585

    2016 doi arXiv

  36. [44]

    Arjovsky, M., Chintala, S., & Bottou, L. (2017). Wasserstein GAN.Proceedings of the 34th International Conference on Machine Learning (ICML), 70, 214-223

    2017

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.