K-Syn: K-space Data Synthesis in Ultra Low-data Regimes

Guan Yu; Liang Dong; Liu Qiegen; Zhang Jianhua

REVIEW 4 major objections 6 minor 23 references

K-Syn: K-space Data Synthesis in Ultra Low-data Regimes

T0 review · 4 major / 6 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read K-Syn claims that a diffusion model trained on 50 cardiac k-space samples, guided by fused amplitude spectra from neighboring frames, generates synthetic data competitive with training on 200 real samples.

desk verdict K-Syn is a plausible k-space synthesis idea with a real internal contradiction and an unvalidated fusion step; it deserves referee time but needs revision. read the letter →

arxiv 2509.06997 v1 pith:CKOWMM3X submitted 2025-09-04 cs.CV

Guan Yu , Zhang Jianhua , Liang Dong , Liu Qiegen This is my paper · ORCID

classification cs.CV

keywords cardiacMRIreconstructionk-spacesynthesislatentdiffusiontemporalfusionultralow-dataregimesfrequency-domaingenerativemodelingdynamicdatascarcity

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

K-Syn is a generative framework for dynamic cardiac MRI that synthesizes raw k-space (frequency-domain) data rather than images. It claims that by randomly fusing amplitude spectra of neighboring time frames while keeping phase fixed, and then generating new latent codes with a diffusion model conditioned on those fused spectra, a model trained on as few as 50 samples produces synthetic k-space data whose quality matches a model trained on 200 real samples. The paper further reports that a reconstruction network trained only on K-Syn synthetic data performs comparably to one trained on real data at 10-fold acceleration. The motivation is practical: if true, generative priors learned from tens of scans can supply the large, diverse training sets that MRI reconstruction models normally require.

What carries the argument

Temporal-fusion guidance: a random convex combination of amplitude spectra from neighboring ECG-gated cardiac frames, K = μ·K_m^A + (1-μ)·K_n^A with μ drawn uniformly from (0,1), while the phase is kept fixed. The fused frequency signal is encoded into a latent code that conditions a U-Net denoiser at every diffusion scale and time step. Coupled with a frequency-domain encoder/decoder that compresses complex k-space volumes into a compact latent space, this makes diffusion tractable for high-resolution multi-frame data and steers generation toward a structure-preserving subspace.

What would settle it

Run the identical K-Syn pipeline but replace temporal-fusion guidance with per-frame amplitude noise of matched L2 magnitude, keeping all other training details fixed. If downstream reconstruction performance (e.g., WKGM-Syn) does not degrade, temporal fusion is not the operative mechanism. A complementary direct test: encode real and fused k-space volumes with K-Syn's trained encoder and compare latent distributions; if fused latents are statistical outliers, the 'feasible subspace' claim is not supported.

Watch

Extended reading notes

Core claim

The central claim is that k-space synthesis in ultra low-data regimes is achievable by moving generative modeling into the frequency domain and exploiting temporal redundancy of cardiac cine. K-Syn treats the Fourier-transformed k-space signal as a global feature space, decomposes it into amplitude and phase, and generates new data by mixing amplitudes of adjacent cardiac frames while leaving phase untouched to preserve anatomy. These fused frequency signals condition a latent diffusion model that learns to denoise in a compressed latent space of k-space volumes. The paper reports that K-Syn trained on 50 real samples yields synthetic data with lower FID, KID, and MMD² than training on 200 r

Load-bearing premise

The method assumes that randomly mixing the amplitude spectra of neighboring cardiac frames, while keeping phase fixed, stays close enough to real k-space data that a diffusion model trained on those mixtures learns a valid prior.

Editorial extensions

If this is right

Training reconstruction models on synthetic k-space data can match training on real data at high acceleration factors, reducing the need for fully sampled acquisitions.
A generative prior learned from tens of cardiac samples can produce diverse, high-fidelity k-space data, directly addressing data scarcity in dynamic MRI.
K-Syn reports lower FID, KID, and MMD² than GAN-based and diffusion-based baselines for cardiac k-space synthesis, indicating that frequency-domain feature-level modeling helps stability in low-data settings.
The temporal-fusion strategy leverages inter-frame redundancy without extra acquisitions, so the method could extend to other dynamic imaging settings where temporal correlation is high.

Reading between the lines

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

The amplitude-mixing principle might transfer to other dynamic modalities with strong temporal structure (e.g., cine CT or fMRI), though phase-fidelity constraints would need to be revalidated per anatomy and contrast.
The claim that fused amplitudes stay on the real-data manifold is not directly measured in the paper; a diagnostic using K-Syn's own encoder to compare latent distributions of real versus fused volumes could reveal when fusion begins to degrade.
Synthetic-data training could be combined with physics-based consistency losses to further close the gap to real-data training, something the paper does not test.
Because the framework decouples compression from generation, K-Syn could be plugged into existing latent diffusion toolchains with minimal changes, letting other frequency-domain priors be swapped in.

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.

Desk Editor's Note

read the letter

Short version: K-Syn is a plausible idea—latent diffusion in k-space with a temporal-fusion guidance that mixes amplitudes across cardiac frames. The direction is genuinely useful for low-data CMR, but the paper's own numbers contradict a headline claim, and the key fusion step is never directly validated. It deserves a referee, not a desk reject, but I would not rely on its results yet.

What is actually new and good: previous work mainly synthesizes images or reconstructs from undersampled data; this letter targets k-space directly and uses the temporal correlation of cine frames as a conditioning signal. That is a legitimate new combination. Empirically, K-Syn beats the unnamed GAN and DM baselines by a large margin on FID/KID/MMD², and in the low-data experiment K-Syn-50 (28.3 FID) does look comparable to Real-200 (32.0), which is the kind of evidence you want for a scarcity claim. The code link is a plus.

Soft spots: - Table II contradicts the text. The text says K-Syn-50/200 achieves lower FID/KID than Real-50/200, but K-Syn-200 (33.8 FID) is worse than Real-200 (32.0). The "K-Syn-50 comparable to Real-200" claim holds, but the broader sentence doesn't. This is a careless error that should have been caught. - Eq. 3 fuses amplitude spectra but keeps phase from one frame. That is not a convex combination of complex k-space data. Where one frame's amplitude is near zero, its phase is arbitrary, so the fused coefficient can have no physical counterpart. The paper asserts this stays near a "feasible subspace" but offers no direct check—no ablation showing fused samples invert to anatomically plausible images, no distributional test on the fused input. Since the whole pipeline depends on this, it is a central unverified assumption. - The downstream reconstruction experiment is incomplete. Section III-D says WKGM-Syn is comparable to WKGM at R=10 but reports no PSNR/SSIM/MSE values. That is the practical payoff of the method, and the numbers are missing. - Implementation details are thin. Baselines are anonymous, and the reported training iterations (350 and 200) seem unusually low for diffusion; either the model and patch size are very small or something is misstated.

Who this is for: researchers working on generative models for raw MRI k-space, and anyone evaluating data-augmentation claims for low-data medical imaging. The core direction is worth keeping an eye on; the evidence is not there yet. I would send it to peer review with a request for a competent stats/medical-imaging referee, and the authors should fix the contradiction, name baselines, report downstream metrics, and add a direct validity check for the fusion step.

Referee Report

4 major / 6 minor

Summary. The paper proposes K-Syn, a framework for synthesizing dynamic cardiac MRI k-space data from very few real samples. The method performs feature-level learning in the frequency domain, uses a temporal-fusion strategy that recombines amplitude spectra of adjacent cardiac frames while keeping phase fixed, and trains a latent diffusion model in a compressed k-space latent space. The authors report that K-Syn outperforms generic GAN- and diffusion-based baselines on FID/KID/MMD^2 (Table I), that K-Syn-50 gives lower FID/KID than Real-200 (Table II), and that a downstream reconstruction model trained solely on K-Syn synthetic k-space data (WKGM-Syn) is qualitatively comparable to one trained on real data (Section III-D). The central claim is that temporal-fusion guidance enables strong generative ability in ultra low-data regimes, so that synthetic k-space data can substitute for real data in training reconstruction models.

Significance. If the central claims hold, the paper addresses an important practical bottleneck in dynamic CMR imaging: the scarcity of fully-sampled k-space data. The direction—performing feature-level modeling directly in the frequency domain and using temporal redundancy as a generative prior—is potentially valuable and goes beyond image-domain synthesis. The paper also provides open-source code and a concrete low-data evaluation protocol, which are strengths. However, the current evidence only partially supports the claims. There is an internal contradiction in Table II, the downstream reconstruction claim is not backed by reported quantitative metrics, and the key temporal-fusion assumption in Eq. (3) is not directly validated. These issues are fixable, but they are load-bearing for the paper's central message.

major comments (4)

[Section III-C, Table II] The text states that 'K-Syn-50/200 achieves lower FID and KID scores compared with Real-50/200', but Table II shows the opposite for the 200-sample setting: K-Syn-200 has FID 33.7968 vs. Real-200 31.9902, and KID 0.0055 vs. 0.0051. Moreover, K-Syn-200 is worse than K-Syn-50 on all three metrics, which is non-monotonic and unexplained. This internal contradiction directly weakens the claim that temporal-fusion helps across dataset sizes. The authors should correct the text, or explain why K-Syn degrades at 200 samples, and provide confidence intervals or repeated-run statistics.
[Section III-D] The claim that WKGM-Syn 'attains quantitative performance comparable to that of WKGM' is unsupported: no PSNR, SSIM, or MSE values are reported anywhere in Section III-D or Fig. 6. Only visual comparisons and error maps are shown. Without the actual numbers, the downstream validity of K-Syn synthetic data—a central practical claim—cannot be assessed. Please report the reconstruction metrics for WKGM-Syn and WKGM, including standard deviations if available.
[Section II-A, Eq. (3)] The temporal-fusion operation combines only amplitude spectra, A_f = μ A_m + (1−μ) A_n, while keeping phase fixed. This is not a convex combination of complex k-space data: k_f = A_f e^{iφ_m} is generally not equal to μ k_m + (1−μ) k_n. At frequencies where frame m has near-zero amplitude, phase is unstable, yet it is inherited by energy taken from frame n. The manuscript asserts that the fused signal 'retains at least as many independent frequency components as any single frame,' but this addresses spectral support, not physical plausibility or proximity to the cardiac k-space manifold. Since the entire method depends on this guidance prior, the authors should provide direct validation, e.g., FID/KID of fused k-space data itself, a reconstruction-based sanity check, or an ablation comparing amplitude-only fusion against complex-domain fusion.
[Section II-C, Eq. (7)] The conditioning mechanism is underspecified. Equation (7) introduces a condition embedding c, and the text says temporal-fusion features encoded as z are 'injected into the model at every scale and time step,' but there is no definition of how c is constructed from the fused latent z, how the encoder output is aggregated, or how conditioning is incorporated in the U-Net. Without this, the method is not reproducible and it is impossible to tell whether the reported benefits come from the fusion prior or from the conditioning architecture. Please provide this implementation detail.

minor comments (6)

[Section II-A, Eq. (1)-(2)] Equation (1) appears garbled: variables h, w, H, W are not clearly defined, and the summation limits are confusing. Equation (2) also has broken notation for the real/imaginary parts. Please rewrite these equations with standard notation.
[Section III-A] The PCA-based selection of 'Real-50/200' and 'K-Syn-50/200' is not described. How many subjects are used, what is the train/test split, and what does PCA selection mean for k-space data? This is important for interpreting the low-data comparison.
[Section III-B, Table I] The baselines are only named 'GAN-based' and 'DM-based' without specific model names or training configurations. Since the comparison is a central contribution, the actual baseline methods should be identified.
[Section III-B] The phrase 'over two times lower' is ambiguous: FID drops from 96.80 to 31.99 (roughly 3x), KID from 0.0172 to 0.0051 (roughly 3.4x). Please give exact ratios or avoid the phrase.
[References] Reference [12] lists the venue as ICML 2025, which is incorrect for Sohl-Dickstein et al. (2015). Please correct this and check all reference metadata.
[Section II-C] The training schedule is very short (350 iterations for the encoder-decoder and 200 iterations for the diffusion model). It would be helpful to state why this is sufficient and whether convergence was observed, especially given the non-monotonic result in Table II.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: K-Syn's fusion-guided diffusion is evaluated against held-out real data rather than fitted to its own outputs.

full rationale

The paper's derivation chain is empirical rather than definitional. Equation (3) defines a temporal-fusion operation A_f = mu*A_m + (1-mu)*A_n while keeping phase intact; this is an input transformation/augmentation used to construct training data and conditioning signals, not a parameter fitted to the evaluation target. The central claims — lower FID/KID/MMD^2 than GAN and diffusion baselines, K-Syn-50 being comparable to Real-200, and WKGM-Syn trained solely on synthetic data being comparable to WKGM — are all tested against real k-space data or against models trained on real data. No equation defines the predicted quantity in terms of the input quantity, and no fitted constant is renamed as a prediction. The paper's few self-citations are standard background references (e.g., Fourier convolution, amplitude/phase interpretation) and are not load-bearing for the main result. The concern that amplitude-only fusion in Eq. (3) may generate off-manifold k-space data is a validity/robustness issue, not a circularity issue: if the fusion prior is implausible, the method would simply underperform or fail, not reduce to its own input. The absence of numerical PSNR/SSIM/MSE values in Section III-D is an incompleteness concern, not a demonstration that the conclusion is forced by construction. Thus no circular step meets the required evidentiary bar.

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

The paper introduces no fitted physical constants and no new entities. The method relies on standard mathematical identities (Fourier transform) and two domain assumptions about amplitude/phase semantics and the validity of temporal-fusion. These assumptions are experimental heuristics rather than derived results.

assumptions (3)

standard math The Fourier transform provides a global receptive field and the convolution theorem holds.
Used in Eq (1)-(2) to justify frequency-domain feature learning; this is a standard mathematical identity.
domain assumption Amplitude and phase of Fourier components correspond to style and structure respectively.
Invoked in Section II-A to justify fusing amplitudes while keeping phases constant; cited to [22] but not proven in this paper.
domain assumption Random convex combinations of amplitude spectra from adjacent frames produce valid and diverse k-space data.
Central to the temporal-fusion strategy in Eq (3) and Section II-A, but no derivation or analysis is given to show these mixtures stay on the true data manifold.

how reviews work

0 comments

Cite this review

Pith. "Pith review of K-Syn: K-space Data Synthesis in Ultra Low-data Regimes." pith.science (2026). https://pith.science/paper/CKOWMM3X

@misc{pith2026250906997,
  author       = {Pith},
  title        = {Pith review of: K-Syn: K-space Data Synthesis in Ultra Low-data Regimes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CKOWMM3X}},
  note         = {Machine review of arXiv:2509.06997}
}

read the original abstract

Owing to the inherently dynamic and complex characteristics of cardiac magnetic resonance (CMR) imaging, high-quality and diverse k-space data are rarely available in practice, which in turn hampers robust reconstruction of dynamic cardiac MRI. To address this challenge, we perform feature-level learning directly in the frequency domain and employ a temporal-fusion strategy as the generative guidance to synthesize k-space data. Specifically, leveraging the global representation capacity of the Fourier transform, the frequency domain can be considered a natural global feature space. Therefore, unlike traditional methods that use pixel-level convolution for feature learning and modeling in the image domain, this letter focuses on feature-level modeling in the frequency domain, enabling stable and rich generation even with ultra low-data regimes. Moreover, leveraging the advantages of feature-level modeling in the frequency domain, we integrate k-space data across time frames with multiple fusion strategies to steer and further optimize the generative trajectory. Experimental results demonstrate that the proposed method possesses strong generative ability in low-data regimes, indicating practical potential to alleviate data scarcity in dynamic MRI reconstruction.

Figures

Figures reproduced from arXiv: 2509.06997 by the authors.

**Figure 1.** Illustration of the difference between (a) pixel-level fusion and (b) feature-level fusion. Typically, the frequency signal is regarded as a special feature descriptor used for fusion. The rapid growth of generative AI has been propelled by several complementary generative model families [6], including variational autoencoders (VAEs) [7], normalizing flows [8], generative adversarial networks (GANs) [9], and denois… view at source ↗

**Figure 2.** Temporal-fusion strategy. Partial per-frame k-space data is fused across adjacent frames to form different frequency feature. B. Latent Representation of K-space Data To enable efficient k-space data generation, we develop a novel compression framework that embeds high-resolution multi-frame k-space volumes into a compact and semantically meaningful latent space. Concretely, a frequency encoder () ñ õ  operates o… view at source ↗

**Figure 3.** Schematic of different temporal-fusion strategies (adjacent, skip, and grouped) jointly guiding latent diffusion process. Once the model is trained, we generate samples by reverse latent diffusion process, starting from ~ (0, ) T z  I and iteratively applying: 1 1 1 ( )( ( ( , , 1 ) )) t t t t t t t t z z z t c ñ           (8) where ~ (0, ) t  I and  t controls the stochasticity. Importantly, temp… view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Qualitative comparison of overall quality in cardiac k-space data synthesis. K-Syn yields sharper myocardial detail and fewer artifacts. C. K-space Data Synthesis in Ultra Low-Data Regimes To evaluate the generative ability of K-Syn in ultra lowdata regimes, we conduc…

**Figure 6.** Figure 6: Reconstruction of the cardiac data at radial sampling of R=10. The second row shows the enlarged view of the ROI region and the third row shows the error map of the reconstruction. IV. CONCLUSION In conclusion, this letter tackles a central bottleneck in dynamic CMR im…

**Figure 5.** Figure 5: Performance comparison of cardiac k-space data synthesis results generated by K-Syn model trained with different numbers of samples. TABLE II QUANTITATIVE RESULTS OF K-SYN MODELS TRAINED WITH DIFFERENT DATASET SIZES. Method FID↓ KID↓ MMD²↓ Real-50 38.1126 0.0061 ± 0.00…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

23 extracted references · 19 canonical work pages

[1]

J. P. Finn, K. Nael, V. Deshpande, et al. , <Cardiac MR imaging: State of the technology,= Radiology, vol. 241, no. 2, pp. 338-354, Nov. 2006

work page 2006
[2]

J. Lyu, C. Qin, S. Wang, et al., <The state-of-the-art in cardiac MRI reconstruction: Results of the CMRrecon challenge in MICCAI 2023,= Med. Image Anal., vol. 101, p. 103485, Apr. 2025

work page 2023
[3]

S. S. Sengar, A. B. Hasan, S. Kumar, et al. , <Generative artificial intelligence: a systematic review and applications,= Multimed. Tools Appl., vol. 84, pp. 23661-23700, Aug. 2024

work page 2024
[4]

D. G. Takale, P. N. Mahalle, and B. Sule, <Advancements and app li- cations of generative artificial intelligence,= J. Inf. Technol. Sci., vol. 10, no. 1, pp. 20-27, Mar. 2024

work page 2024
[5]

Trigka and E

M. Trigka and E. Dritsas, <The evolution of generative AI: Tren ds and applications,= IEEE Access, May. 2025

work page 2025
[6]

Bengesi, H

S. Bengesi, H. El-Sa yed, M. K. Sarker, et al. , <Advancements in generative AI: A comprehensive review of GANs, GPT, autoencod- ers, diffusion model, and transformers,= IEEE Access , vol. 12, pp. 69812-69837, 2024

work page 2024
[7]

D. P. Kingma and M. Welling, <Au to-encoding variational bayes,= arXiv preprint arXiv:1312.6114, 2013

arXiv 2013
[8]

Rezende and S

D. Rezende and S. Mohamed, <Vari ational inference with normaliz - ing flows,= in Proc. Int. Conf. Mach. Learn. , Lille, France, 2015, pp. 1530-1538

work page 2015

Show all 23 references

[9]

Goodfellow, J

I. Goodfellow, J. Pouget-Abadie, M. Mirza, et al. , <Generative ad- versarial networks,= Commun. ACM, vol. 63, no. 11, pp. 139-144, Oct. 2020

2020
[10]

J. Ho, A. Jain, and P. Abbeel, <Denoising diffusion probabilist ic models,= in Proc. Adv. Neural Inf. Process. Syst. ,2020, pp. 6840- 6851

2020
[11]

T. Che, Y. Li and W. Li, <Mode r egularized generative adversari al networks,= arXiv preprint arXiv:1612.02136, 2016

2016 arXiv
[12]

Sohl-Dickstein, E

J. Sohl-Dickstein, E. Weiss and S. Ganguli, <Deep unsupervised learning using nonequilibrium thermodynamics,= in Proc. Int. Conf. Mach. Learn., Lille, France, 2025, pp. 2256-2265

2025
[13]

Rombach, A

R. Rombach, A. Blattmann, D. Lorenz, et al., <High-resolution im- age synthesis with latent diffusion models,= in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit.,2022, pp. 10684-10695

2022
[14]

W a n g , X

W . W a n g , X . Y u , C . W a n g , et al. , <One for multiple: Physics- informed synthetic data boosts generalizable deep learning for fast MRI reconstruction,= Med. Image Anal., vol. 103, p. 103616, 2025

2025
[15]

Friedrich, J

P. Friedrich, J. Wolleb and F. Bieder, <Wdm: 3D wavelet diffusi on models for high-resolution medical image synthesis,= in Proc. Int. Conf. Med. Image Comput. Comput.Assist. Intervent. ,Switzerland, 2024, pp. 11-21

2024
[16]

H. Wang, Z. Liu, K. Sun, et al., <3D MedDiffusion: A 3D medical latent diffusion model for cont rollable and high-quality medica l im- age generation,= IEEE Trans. Med. Imaging, 2025

2025
[17]

F. Atli, B. Kabas, F. Arslan, et al., <I2I-Mamba: Multi-modal medi- cal image synthesis via selective state space modeling,= arXiv pre- print arXiv:2405.14022, May. 2024

2024
[18]

Friedrich, A

P. Friedrich, A. Durrer, J. Wolleb, et al. , <CWDM: Conditional wavelet diffusion models for cro ss-modality 3D medical image sy n- thesis,= arXiv preprint arXiv:2411.17203, Nov. 2024

2024 arXiv
[19]

S. U. R. Khan and S. Asif, <Oral cancer detection using feature -level fusion and novel self-attention mechanisms,= Biomed. Signal Pro- cess. Control, vol. 95, p. 106437, 2024

2024
[20]

Atrey, B

K. Atrey, B. K. Singh, and N. K. Bodhey, <Multimodal classifica tion of breast cancer using feature level fusion of mammogram and ul tra- sound images in machine learning paradigm,= Multimed. Tools Appl., vol. 83, no. 7, pp. 21347-21368, 2024

2024
[21]

H. Sun, Y. Li, Z. Li, et al., <Fourier convolution block with global receptive field for MRI reconstruction,= Med. Image Anal., vol. 99, p. 103349, Jan. 2025

2025
[22]

J. Xian, X. Li and D. Tu, <Unsupervised cross-modality adaptati on via dual structural-oriented guidance for 3D medical image segm en- tation,= IEEE Trans. Med. Imaging , vol. 42, no.6, pp. 1774-1785, 2023

2023
[23]

Z. Yue, J. Wang, C. C. Loy, <Efficient diffusion model for imag e restoration by residual shifting,= IEEE Trans. Pattern Anal. Mach. Intell., 2024

2024

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

J. P. Finn, K. Nael, V. Deshpande, et al. , <Cardiac MR imaging: State of the technology,= Radiology, vol. 241, no. 2, pp. 338-354, Nov. 2006

work page 2006

[2] [2]

J. Lyu, C. Qin, S. Wang, et al., <The state-of-the-art in cardiac MRI reconstruction: Results of the CMRrecon challenge in MICCAI 2023,= Med. Image Anal., vol. 101, p. 103485, Apr. 2025

work page 2023

[3] [3]

S. S. Sengar, A. B. Hasan, S. Kumar, et al. , <Generative artificial intelligence: a systematic review and applications,= Multimed. Tools Appl., vol. 84, pp. 23661-23700, Aug. 2024

work page 2024

[4] [4]

D. G. Takale, P. N. Mahalle, and B. Sule, <Advancements and app li- cations of generative artificial intelligence,= J. Inf. Technol. Sci., vol. 10, no. 1, pp. 20-27, Mar. 2024

work page 2024

[5] [5]

Trigka and E

M. Trigka and E. Dritsas, <The evolution of generative AI: Tren ds and applications,= IEEE Access, May. 2025

work page 2025

[6] [6]

Bengesi, H

S. Bengesi, H. El-Sa yed, M. K. Sarker, et al. , <Advancements in generative AI: A comprehensive review of GANs, GPT, autoencod- ers, diffusion model, and transformers,= IEEE Access , vol. 12, pp. 69812-69837, 2024

work page 2024

[7] [7]

D. P. Kingma and M. Welling, <Au to-encoding variational bayes,= arXiv preprint arXiv:1312.6114, 2013

arXiv 2013

[8] [8]

Rezende and S

D. Rezende and S. Mohamed, <Vari ational inference with normaliz - ing flows,= in Proc. Int. Conf. Mach. Learn. , Lille, France, 2015, pp. 1530-1538

work page 2015