Systematic comparison of deep generative models applied to multivariate financial time series

Howard Caulfield; James P. Gleeson

REVIEW 3 major objections 6 minor 56 references

Systematic comparison of deep generative models applied to multivariate financial time series

T0 review · 3 major / 6 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 argues that deep generative models—led by RCGAN—can add value in multivariate financial return modelling, beating parametric baselines on synthetic data and improving volatility-trading signals on empirical data.

desk verdict Useful synthetic benchmark for multivariate return generators, but the economic-value claim rests on an underspecified and likely misspecified HAR-with-generated-features trading signal. read the letter →

arxiv 2412.06417 v1 pith:XJM4WXHI submitted 2024-12-09 q-fin.ST q-fin.CP

Howard Caulfield , James P. Gleeson This is my paper · ORCID

classification q-fin.STq-fin.CP

keywords deepgenerativemodelsfinancialtimeseriesmultivariatereturnsRCGANGMMNHARmodelvolatilitytradingEarthMover'sDistance

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 claims that deep generative models (DGMs) can add value in multivariate financial return modelling, going beyond what standard parametric volatility models achieve. It supports this through a systematic comparison: conditional DGMs such as RCGAN, GMMN, TimeGAN, CoMeTS, CTVAE, and CTNF compete against DCC-GARCH, Copula GARCH, and factor stochastic volatility on synthetic datasets of increasing complexity (NGARCH, Heston, and their regime/jump variants). The headline result is that RCGAN ranks best overall on the hardest datasets, with an average rank of 4.80, and that volatility features generated by RCGAN and GMMN improve the profit-and-loss of a HAR-based straddle-basket trading strategy relative to the HAR baseline. If right, this suggests generative models can serve as flexible, assumption-light replacements or complements for parametric return generators in risk management and portfolio applications.

What carries the argument

The central machinery is a conditional generation setup built on the AR-FNN (autoregressive feedforward neural network) architecture, where each time-step output is a function of a rolling window of past returns plus a noise vector, enabling generation of arbitrary length series. RCGAN is a recurrent conditional GAN using this architecture; GMMN is a moment-matching network trained with maximum mean discrepancy, extended to include absolute-return and correlation losses. Evaluation uses Earth Mover's Distance between true and generated distributions of mean, standard deviation, skew, kurtosis, and correlations, over both full series and rolling windows. The empirical task substitutes expected future daily, weekly, and monthly realized volatility from generated batches into the HAR realized-volatility model, then ranks instruments by predicted-volatility-to-implied-volatility ratios to build theta-neutral straddle baskets.

What would settle it

Run the same HAR-plus-generated-features pipeline but compute the straddle PnL with the full bid-ask spread and include vega profit and transaction fees, then bootstrap the per-day PnL differences between generative-HAR and baseline HAR; if the 95% bootstrap confidence interval for the difference includes zero, or the sign flips under realistic costs, the claim that DGMs add value in trading would be unsupported.

Watch

Extended reading notes

Core claim

The central discovery is that implicit deep generative models—particularly the recurrent conditional GAN RCGAN—can capture multivariate financial return distributions as well as or better than state-of-the-art parametric models specified to match those distributions. On NGARCH+ data, RCGAN achieves the lowest Earth Mover's Distances across moment, correlation, and rolling-window measures; on Heston+ data, no model dominates, but RCGAN and GMMN are the strongest DGMs and rank competitively with the best parametric alternatives. When the generated returns are used to construct HAR volatility features for a theta-neutral straddle basket of S&P 500 constituents, the generative features produce higher profit-per-day than the HAR baseline on long/short, long-only, and short-only baskets. The authors interpret this as evidence that DGMs can add value in multivariate financial return modelling and could act as foundation models for economic applications.

Load-bearing premise

The entire empirical demonstration rests on the assumption that the simulated straddle-basket PnL, computed with three-quarter-spread approximations and excluding vega profit and transaction fees, is a reliable measure of real trading performance; if that approximation is too crude, the reported generative-HAR outperformance could vanish under realistic costs.

Editorial extensions

If this is right

If the central claim is right, generative return models can serve as foundation models: pretrained conditional generators whose features improve downstream volatility forecasting and trading signals.
The ranking result suggests that simple implicit models like RCGAN and GMMN may be enough to capture multivariate return distributions, sidestepping explicit priors such as Gaussian copulas or factor structures.
The improved PnL of generative-HAR over baseline HAR implies that generated conditional distributions encode predictive information about future realized volatility beyond what the HAR's lagged volatility terms capture.
The negative result that network features based on generated correlations add no value—and that neither DGM captures empirical dynamic correlation—points to a concrete limitation: current DGMs are strong marginal generators but weak conditional copula learners.
The success on Heston+ with jumps and regimes suggests the approach may extend to realistic data, though the empirical Jaccard-index analysis tempers this for dynamic correlation.

Reading between the lines

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

The PnL gaps between generative-HAR and baseline HAR are reported as point values without confidence intervals; a bootstrap over test days would show whether the outperformance is distinguishable from noise, and adding realistic transaction costs and vega exposure could erode it.
The paper's framing suggests a natural next experiment: test RCGAN and GMMN features as inputs to already-established volatility models (for example, GARCH-family or higher-frequency HAR variants) to see whether the gain persists across horizons and asset classes.
The failure to learn dynamic correlation in the empirical data hints that a graph-aware generator—one that conditions on a learned adjacency matrix—might capture the network effects the HAR baseline already exploits; this would be a direct testable extension of the paper's framework.
Because the synthetic comparison rewards models that match unconditional moments, the ranking may overstate usefulness for conditional risk applications; the authors' own Jaccard-index analysis is a partial admission of this gap.

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

Dear [Colleague],

The paper is a systematic comparison of deep generative models (RCGAN, GMMN, TimeGAN, CoMeTS, CTVAE, CTNF) against parametric multivariate models (FSV, DCC, Copula GARCH) for multivariate financial return generation. The new content is the evaluation protocol: synthetic datasets with increasing complexity (NGARCH, Heston, plus versions with regimes and jumps) and a range of distribution-distance measures on moments, correlations, and rolling windows, followed by an empirical implied-volatility trading task. The synthetic benchmark is honestly built and the results are plausible: RCGAN ranks first overall, with GMMN and FSVR close behind. This is a useful reference point for practitioners choosing generators.

The paper does well on transparency in its own limitations—it flags FSV covariance failures, DGM artifacts, and the exclusion of diffusion models—and the conclusion is measured. The synthetic comparison alone is a legitimate contribution.

The soft spot is the empirical section. The trading-signal construction in Section 4.3 is one sentence: 'we take the expected future daily, weekly and monthly realized volatility over all generated batches. We substitute these features into the baseline HAR model.' This is underspecified and the most natural reading is a misspecification: you would be regressing future RV on its own forecast using coefficients estimated for lagged RVs, or else comparing a direct DGM forecast to a HAR forecast without a common framework. Either way, the PnL figures in Figures 1-3 may not measure what the paper claims. This is load-bearing because the paper frames the trading task as the demonstration of DGM value. The reader's concern about missing error bars, costs, and vega is real but secondary; even a perfectly costed PnL would be meaningless if the signal construction is conceptually flawed. No code or pseudocode is released, so the procedure can't be checked.

Minor issues: the equal-weight rank aggregation in Table 6 is arbitrary and has no sensitivity analysis; the FSV rolling results are conditioned on valid covariance draws, a selection the authors acknowledge; and the no-code-release policy limits reproducibility.

Who is this for? Someone shopping for a multivariate return generator or working on financial GAN benchmarks. It deserves a serious referee: the synthetic comparison is careful and the trading application, once clarified, could be a valuable case study. My recommendation: send it to review, but the authors should re-estimate or properly specify the HAR-with-generated-features model, add significance tests and costs to the PnL, and release code. If the empirical section can't be fixed, the paper would still stand on the synthetic benchmark alone.

Regards,

[Your name]

Referee Report

3 major / 6 minor

Summary. The paper systematically compares six deep generative models (RCGAN, TimeGAN, GMMN, CoMeTS, CTVAE, CTNF) against parametric factor stochastic volatility and multivariate GARCH variants on synthetic NGARCH+ and Heston+ datasets, using Earth Mover's Distances of moment and correlation distributions over full and rolling windows. On the synthetic benchmarks, RCGAN has the best overall average rank (4.80 in Table 6), with GMMN also competitive. The authors then apply RCGAN and GMMN to an empirical options dataset, construct a theta-neutral straddle basket whose signals come from HAR realized-volatility forecasts augmented with generated features (Section 4.3), and report that the generative HAR models outperform the baseline HAR in daily PnL (Figures 1-3). The paper concludes that DGMs can add value in multivariate financial return modelling, primarily on the strength of this empirical trading task.

Significance. The synthetic comparison is a useful and fairly broad benchmark: it uses increasingly complex synthetic datasets, compares implicit and explicit DGMs with parametric baselines, evaluates correlation and rolling-window moment distances, and averages over five seeds. The authors also include an honest analysis of where the best DGM fails (e.g., rolling standard deviation bimodality, dynamic correlation). If the empirical trading claim were established, the paper would make a solid contribution to the q-fin.ST literature. However, the headline empirical conclusion currently rests on an underspecified 'generative HAR' construction and on point PnL figures without uncertainty quantification or transaction costs; as written, the economic claim is not yet convincing.

major comments (3)

[Section 4.3, Eq. (9)] The construction of the generative HAR features is too terse to support the central PnL claim. The text says: 'we take the expected future daily, weekly and monthly realized volatility over all generated batches. We substitute these features into the baseline HAR model.' If those features are expected future RVs computed from generated future paths, they are forecasts of the dependent variable, not the lagged RV regressors in Eq. (9); substituting them into a HAR model whose coefficients were estimated on lagged RVs is a misspecification, and the comparison with the baseline is not like-for-like. If instead the generated features are used as direct forecasts, the benchmark should be a direct HAR forecast rather than the recursive HAR in Eq. (9), and the source of any PnL improvement is unclear. Please specify the exact timing: the conditioning window, the generated horizon, how features are aggregated across generated batches and seeds, and whether the HAR coefficients are re-estimated on the generated features or applied unchanged. Also state explicitly every step that ensures no look-ahead. Without this, Figures 1-3 do not measure what the paper claims.
[Section 5.2, Figures 1-3] The PnL results are reported as point values without error bars, confidence intervals, or significance tests. With only five seeds and a single empirical test period, the claimed 'clear outperformance' and 'stark' differences could be sampling noise. The PnL also relies on a three-quarter-spread approximation and explicitly excludes vega profit and transaction fees (Sections 3.2 and 4.3). Please report variability across seeds and time (e.g., block bootstrap or subperiod analysis), and show how the conclusions change under alternative spread, fee, and vega assumptions. The exclusion of costs is particularly important because the economic claim is about adding value in trading.
[Section 4.1 and Table 6] The ranking that supports 'RCGAN is the clear best performer' (Section 5.1) uses an arbitrary equal-weighted average over ten distance measures in Table 6. The paper does not report the variance of these ranks across the five seeds or the sensitivity of the combined ranking to the aggregation scheme. Please report per-seed ranks and test alternative aggregations (e.g., median rank, worst-case rank, or separate per-dataset rankings). Without this robustness check, the headline ranking may not be stable.

minor comments (6)

[Abstract and Section 1] The phrase 'a implied volatility trading task' should be 'an implied volatility trading task'.
[Section 3.3] The sentence 'The we examine are mean, standard deviation, skew and kurtosis' is incomplete; it should read something like 'The measures we examine are...'.
[Section 4.1] The formatting 'rpackages factorstochvol [24] and rmgarch [21]' should be 'R packages factorstochvol [24] and rmgarch [21]'.
[Figures 1-3 captions] The captions of Figures 1, 2, and 3 refer to 'This table represents the profit per day...' but these are figures, not tables; please correct the wording.
[Section 5.2] The text refers to 'see figure 4 for possible reasons' before Figure 4 is described; consider moving or rephrasing to make the cross-reference clearer.
[Section 5.1] Table 6's caption says 'Columns are sorted based on ascending combined rank', but the table lists rows in that order; please clarify whether the ordering is by row or column.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: DGM comparisons use external synthetic and empirical targets, and the HAR baseline is an independent benchmark.

full rationale

The paper's derivation chain is self-contained against external benchmarks. Synthetic-data evaluation compares generated samples to the true NGARCH+/Heston+ simulation outputs through Earth Mover's Distance on moment and rolling-moment distributions; the targets are not constructed from the generated outputs. The empirical application uses a standard HAR realized-volatility model (Eq. 9) with coefficients estimated from realized volatility data, and the DGM-based HAR variant replaces lagged volatility regressors with features obtained from generated future paths. Even if this substitution is econometrically debatable as a forecasting specification, it is not circular: the DGM features are model outputs, not fitted parameters calibrated to the PnL target, and the baseline HAR is independently estimated. Hyperparameters are selected on validation distribution-distance metrics that resemble the evaluation metrics, which is a model-selection overlap rather than a derivation-level equivalence; no fitted parameter is renamed as a prediction. The paper contains no load-bearing self-citations: the cited prior works are by other authors and are used for architecture and baseline methods, not to justify the central claim. The PnL figures lack confidence intervals and rely on Greek approximations, but those are statistical robustness concerns, not circularity. No equation in the paper is equivalent to its inputs by construction, and no claim reduces to a self-citation chain.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The benchmark depends on many hand-set choices: synthetic process parameters and regime or jump settings are not reported, the conditioning window is fixed at 40, the ranking uses equal weights over ten EMD measures, and the empirical trading setup uses approximate straddle prices and exclusion rules. No new physical or model entity is introduced. The paper supplies no code or data, so these choices cannot be audited externally.

free parameters (6)

DGM hyperparameters = not reported
Selected by grid search on the validation set in Section 4.1; exact values are not given, and they drive all DGM results.
Conditioning window length = 40 time steps
Used for all conditioned models and the rolling parametric baselines in Section 4.1; no sensitivity analysis is reported.
Synthetic regime and jump parameters = not reported
NGARCH+ and Heston+ datasets are built from hand-chosen parameterizations and regime or jump settings in Section 3.2; exact values are not listed.
Equal-weight ranking weights = equal weights
Table 6 uses a naive average rank over all EMD measures in Section 3.3; the equal weighting is chosen by hand.
Straddle pricing assumptions = 3/4 bid-ask spread, 50 bp strike rule
Section 4.3 approximates straddle prices and excludes trades using these values; no sensitivity tests are reported.
HAR ridge and exponential weighting parameters = not reported
The baseline HAR model uses ridge regression and exponential weighting in Section 4.3 with unstated penalty and decay values.

assumptions (5)

domain assumption NGARCH(1,1) and multi-asset Heston processes, plus regime and jump extensions, capture the features that matter for benchmarking multivariate financial return generators.
The synthetic datasets defined in Section 3.2 and Table 2 are the primary basis for model ranking in Section 5.1; if these synthetic processes are not representative of real markets, the ranking may not transfer.
domain assumption Earth Mover's Distance over marginal moment distributions and correlation distributions is a sufficient measure of generation quality.
Section 3.3 uses no other evaluation metric for synthetic data; path dependencies beyond rolling moments are not scored.
domain assumption The R implementations of DCC, Copula-GARCH, and factor stochastic volatility are correctly used and are appropriate state-of-the-art baselines.
Section 4.1 relies on the rmgarch and factorstochvol packages for all parametric results in Tables 4-6.
ad hoc to paper Plugging DGM-generated future realized volatility features into a HAR model fitted by ridge regression with exponential weighting does not introduce look-ahead bias.
Section 4.3 uses this substitution to build trading signals; the paper does not verify that the HAR coefficients remain valid on generated features.
domain assumption FSV MCMC chains converge with the stated burn-in and draw counts.
Section 4.1 uses 500 burn-in and 5000 draws with thinning 5; no convergence diagnostics are reported.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Systematic comparison of deep generative models applied to multivariate financial time series." pith.science (2026). https://pith.science/paper/XJM4WXHI

@misc{pith2026241206417,
  author       = {Pith},
  title        = {Pith review of: Systematic comparison of deep generative models applied to multivariate financial time series},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XJM4WXHI}},
  note         = {Machine review of arXiv:2412.06417}
}

read the original abstract

Financial time series (FTS) generation models are a core pillar to applications in finance. Risk management and portfolio optimization rely on realistic multivariate price generation models. Accordingly, there is a strong modelling literature dating back to Bachelier's Theory of Speculation in 1901. Generating FTS using deep generative models (DGMs) is still in its infancy. In this work, we systematically compare DGMs against state-of-the-art parametric alternatives for multivariate FTS generation. We initially compare both DGMs and parametric models over increasingly complex synthetic datasets. The models are evaluated through distance measures for varying distribution moments of both the full and rolling FTS. We then apply the best performing DGM models to empirical data, demonstrating the benefit of DGMs through a implied volatility trading task.

Figures

Figures reproduced from arXiv: 2412.06417 by the authors.

**Figure 4.** The y-axis represents the average Jaccard index of the future correlation network to the past and model generated correlation networks for all time steps in the test dataset. The x-axis represents the percentile for the correlation threshold i.e., 90 represents a correlation network based on the 90𝑡ℎ largest percentile correlation value for a sliding window of length 40. A value of one implies perfect match. The b… view at source ↗

**Figure 3.** Similar to figure 1, this table shows the PnL from [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 6.** Figure 6: While successfully learning the majority of return characteristics some artifacts still exist. In this histogram, we see the frequencies of rolling standard deviation values for each time series generated. The best model has trouble learning the bi-modality (second b…

**Figure 7.** Figure 7: The histograms represent the normalized count of correlation values for the synthetic (red) and true (blue) data. The figure on the left shows the correlation structure of the synthetic returns early in the training phase. After training the figure on the right shows …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 44 canonical work pages

[1]

Martin Arjovsky, Soumith Chintala, and Léon Bottou. 2017. Wasserstein genera- tive adversarial networks. In International conference on machine learning . PMLR, 214–223

2017
[2]

Louis Bachelier. 1901. Théorie mathématique du jeu. In Annales Scientifiques de l’Ecole Normale Supérieure, Vol. 18. 143–209

work page 1901
[3]

Luc Bauwens, Sébastien Laurent, and Jeroen VK Rombouts. 2006. Multivariate GARCH models: a survey. Journal of applied econometrics 21, 1 (2006), 79–109

work page 2006
[4]

Christopher M Bishop. 1994. Mixture density networks. (1994)

work page 1994
[5]

Tim Bollerslev. 1990. Modelling the coherence in short-run nominal exchange rates: a multivariate generalized ARCH model. The review of economics and statistics (1990), 498–505

work page 1990
[6]

Nicolas Bonneel, Michiel Van De Panne, Sylvain Paris, and Wolfgang Heidrich

work page
[7]

Qinkai Chen and Christian-Yann Robert. 2022. Multivariate realized volatility forecasting with graph neural network. In Proceedings of the Third ACM interna- tional Conference on AI in Finance . 156–164

work page 2022
[8]

Yingmei Chen, Zhongyu Wei, and Xuanjing Huang. 2018. Incorporating Corpo- ration Relationship via Graph Convolutional Neural Networks for Stock Price Prediction. Proceedings of the 27th ACM International Conference on Information and Knowledge Management (2018)

work page 2018

Show all 56 references

[9]

Dawei Cheng, Fangzhou Yang, Sheng Xiang, and Jin Liu. 2022. Financial time series forecasting with multi-modality graph neural network. Pattern Recognition 121 (2022), 108218

2022
[10]

Adam Clements and Daniel PA Preve. 2021. A practical guide to harnessing the HAR volatility model. Journal of Banking & Finance 133 (2021), 106285

2021
[11]

Rama Cont. 2001. Empirical properties of asset returns: stylized facts and statisti- cal issues. Quantitative finance 1, 2 (2001), 223

2001
[12]

Rama Cont, Mihai Cucuringu, Renyuan Xu, and Chao Zhang. 2022. Tail-gan: Nonparametric scenario generation for tail risk estimation. arXiv preprint arXiv:2203.01664 (2022)

2022 arXiv
[13]

Fulvio Corsi. 2009. A simple approximate long-memory model of realized volatil- ity. Journal of Financial Econometrics 7, 2 (2009), 174–196

2009
[14]

Fernando de Meer Pardo, Peter Schwendner, and Marcus Wunsch. 2022. Tackling the Exponential Scaling of Signature-Based Generative Adversarial Networks for High-Dimensional Financial Time-Series Generation. The Journal of Financial Data Science 4, 4 (2022), 110–132

2022
[15]

Georgi Dimitroff, Stefan Lorenz, and Alexander Szimayer. 2011. A parsimonious multi-asset Heston model: Calibration and derivative pricing. International Journal of Theoretical and Applied Finance 14, 08 (2011), 1299–1333

2011
[16]

Laurent Dinh, Jascha Sohl-Dickstein, and Samy Bengio. 2022. Density estimation using Real NVP. In International Conference on Learning Representations

2022
[17]

Mihai Dogariu, Liviu-Daniel S, tefan, Bogdan Andrei Boteanu, Claudiu Lamba, Bomi Kim, and Bogdan Ionescu. 2022. Generation of Realistic Synthetic Financial Time-series. ACM Transactions on Multimedia Computing, Communications, and Systematic comparison of deep generative model...

2022
[18]

Robert Engle. 2002. Dynamic conditional correlation: A simple class of multivari- ate generalized autoregressive conditional heteroskedasticity models. Journal of business & economic statistics 20, 3 (2002), 339–350

2002
[19]

Lars Ericson, Xuejun Zhu, Xusi Han, Rao Fu, Shuang Li, Steve Guo, and Ping Hu
[20]

Cristóbal Esteban, Stephanie L Hyland, and Gunnar Rätsch. 2017. Real-valued (medical) time series generation with recurrent conditional gans. arXiv preprint arXiv:1706.02633 (2017)

2017 arXiv
[21]

Alexios Galanos. 2022. rmgarch: Multivariate GARCH models. R package version 1.3-9.0

2022
[22]

Federico Gatta, Fabio Giampaolo, Edoardo Prezioso, Gang Mei, Salvatore Cuomo, and Francesco Piccialli. 2022. Neural networks generative models for time series. Journal of King Saud University-Computer and Information Sciences 34, 10 (2022), 7920–7939

2022
[23]

Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. 2020. Generative adversarial networks. Commun. ACM 63, 11 (2020), 139–144

2020
[24]

Darjus Hosszejni and Gregor Kastner. 2021. Modeling Univariate and Multivariate Stochastic Volatility in R with stochvol and factorstochvol. Journal of Statistical Software 100, 12 (2021), 1–34. https://doi.org/10.18637/jss.v100.i12

2021 doi
[25]

Eric Jondeau and Michael Rockinger. 2006. The copula-garch model of conditional dependencies: An international stock market application. Journal of international money and finance 25, 5 (2006), 827–853

2006
[26]

Gregor Kastner, Sylvia Frühwirth-Schnatter, and Hedibert Freitas Lopes. 2017. Efficient Bayesian inference for multivariate factor stochastic volatility models. Journal of Computational and Graphical Statistics 26, 4 (2017), 905–917

2017
[27]

Diederik P Kingma and Max Welling. 2014. Auto-encoding variational Bayes. In International Conference on Learning Representations

2014
[28]

Alexei Kondratyev and Christian Schwarz. 2019. The market generator.A vailable at SSRN 3384948 (2019)

2019
[29]

Alireza Koochali, Peter Schichtel, Andreas Dengel, and Sheraz Ahmed. 2019. Probabilistic forecasting of sensory data with generative adversarial networks– forgan. IEEE Access 7 (2019), 63868–63880

2019
[30]

Colin Lea, Michael D Flynn, Rene Vidal, Austin Reiter, and Gregory D Hager
[31]

Greg Lewis and Vasilis Syrgkanis. 2018. Adversarial generalized method of moments. arXiv preprint arXiv:1803.07164 (2018)

2018 arXiv
[32]

Yujia Li, Kevin Swersky, and Rich Zemel. 2015. Generative moment matching networks. In International conference on machine learning . PMLR, 1718–1727

2015
[33]

Terry Lyons. 2014. Rough paths, signatures and the modelling of functions on streams. arXiv preprint arXiv:1405.4537 (2014)

2014 arXiv
[34]

Giovanni Mariani, Yada Zhu, Jianbo Li, Florian Scheidegger, Roxana Istrate, Costas Bekas, and A Cristiano I Malossi. 2019. Pagan: Portfolio analysis with generative adversarial networks. arXiv preprint arXiv:1909.10578 (2019)

2019 arXiv
[35]

Giuseppe Masi, Matteo Prata, Michele Conti, Novella Bartolini, and Svitlana Vyetrenko. 2023. On Correlated Stock Market Time Series Generation.Proceedings of the Fourth ACM International Conference on AI in Finance (2023)

2023
[36]

Hao Ni, Lukasz Szpruch, Marc Sabate-Vidales, Baoren Xiao, Magnus Wiese, and Shujian Liao. 2021. Sig-Wasserstein GANs for time series generation. In Proceedings of the Second ACM International Conference on AI in Finance . 1–8

2021
[37]

Hao Ni, Lukasz Szpruch, Magnus Wiese, Shujian Liao, and Baoren Xiao. 2020. Conditional Sig-Wasserstein GANs for Time Series Generation. DecisionSciRN: Probabilistic Graphical Models (Topic) (2020)

2020
[38]

Danilo Rezende and Shakir Mohamed. 2015. Variational inference with normaliz- ing flows. In International Conference on Machine Learning . PMLR, 1530–1538

2015
[39]

Annastiina Silvennoinen and Timo Teräsvirta. 2009. Multivariate GARCH models. In Handbook of financial time series . Springer, 201–229

2009
[40]

Euan Sinclair. 2013. Volatility trading. John Wiley & Sons

2013
[41]

Justin Sirignano and Rama Cont. 2021. Universal features of price formation in financial markets: perspectives from deep learning. In Machine learning and AI in finance. Routledge, 5–15

2021
[42]

Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Ste- fano Ermon, and Ben Poole. [n. d.]. Score-Based Generative Modeling through Stochastic Differential Equations. In International Conference on Learning Repre- sentations
[43]

Yang Song, Jascha Sohl-Dickstein, Diederik P Kingma, Abhishek Kumar, Stefano Ermon, and Ben Poole. 2020. Score-based generative modeling through stochastic differential equations. arXiv preprint arXiv:2011.13456 (2020)

2020 arXiv
[44]

Shuntaro Takahashi, Yu Chen, and Kumiko Tanaka-Ishii. 2019. Modeling financial time-series with generative adversarial networks. Physica A: Statistical Mechanics and its Applications (2019)

2019
[45]

Ruslan Tepelyan and Achintya Gopal. 2023. Generative Machine Learning for Multivariate Equity Returns. Proceedings of the Fourth ACM International Confer- ence on AI in Finance (2023)

2023
[46]

Yiu Kuen Tse and Albert K C Tsui. 2002. A multivariate generalized autoregressive conditional heteroscedasticity model with time-varying correlations. Journal of Business & Economic Statistics 20, 3 (2002), 351–362

2002
[47]

Benigno Uria, Marc-Alexandre Côté, Karol Gregor, Iain Murray, and Hugo Larochelle. 2016. Neural autoregressive distribution estimation. The Journal of Machine Learning Research 17, 1 (2016), 7184–7220

2016
[48]

Milena Vuletić, Felix Prenzel, and Mihai Cucuringu. 2024. Fin-gan: Forecast- ing and classifying financial time series via generative adversarial networks. Quantitative Finance 24, 2 (2024), 175–199

2024
[49]

Yuanrong Wang and Tomaso Aste. 2022. Network Filtering of Spatial-temporal GNN for Multivariate Time-series Prediction. In Proceedings of the Third ACM International Conference on AI in Finance . 463–470

2022
[50]

Yuanrong Wang, Antonio Briola, and Tomaso Aste. 2023. Topological Portfo- lio Selection and Optimization. In Proceedings of the Fourth ACM International Conference on AI in Finance . 681–688

2023
[51]

Magnus Wiese, Robert Knobloch, Ralf Korn, and Peter Kretschmer. 2020. Quant GANs: deep generation of financial time series. Quantitative Finance 20, 9 (2020), 1419–1440

2020
[52]

Sheng Xiang, Dawei Cheng, Chencheng Shang, Ying Zhang, and Yuqi Liang. 2022. Temporal and Heterogeneous Graph Neural Network for Financial Time Series Prediction. Proceedings of the 31st ACM International Conference on Information & Knowledge Management (2022)

2022
[53]

Jinsung Yoon, Daniel Jarrett, and Mihaela van der Schaar. 2019. Time-series Generative Adversarial Networks. In Neural Information Processing Systems

2019
[2011]

InProceedings of the 2011 SIGGRAPH Asia conference

Displacement interpolation using Lagrangian mass transport. InProceedings of the 2011 SIGGRAPH Asia conference . 1–12

2011
[2017]

In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition

Temporal convolutional networks for action segmentation and detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 156–165
[2024]

ArXiv abs/2401.10370 (2024)

Deep Generative Modeling for Financial Time Series with Application in VaR: A Comparative Review. ArXiv abs/2401.10370 (2024)

2024 arXiv

Pith tools

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

[1] [1]

Martin Arjovsky, Soumith Chintala, and Léon Bottou. 2017. Wasserstein genera- tive adversarial networks. In International conference on machine learning . PMLR, 214–223

2017

[2] [2]

Louis Bachelier. 1901. Théorie mathématique du jeu. In Annales Scientifiques de l’Ecole Normale Supérieure, Vol. 18. 143–209

work page 1901

[3] [3]

Luc Bauwens, Sébastien Laurent, and Jeroen VK Rombouts. 2006. Multivariate GARCH models: a survey. Journal of applied econometrics 21, 1 (2006), 79–109

work page 2006

[4] [4]

Christopher M Bishop. 1994. Mixture density networks. (1994)

work page 1994

[5] [5]

Tim Bollerslev. 1990. Modelling the coherence in short-run nominal exchange rates: a multivariate generalized ARCH model. The review of economics and statistics (1990), 498–505

work page 1990

[6] [6]

Nicolas Bonneel, Michiel Van De Panne, Sylvain Paris, and Wolfgang Heidrich

work page

[7] [7]

Qinkai Chen and Christian-Yann Robert. 2022. Multivariate realized volatility forecasting with graph neural network. In Proceedings of the Third ACM interna- tional Conference on AI in Finance . 156–164

work page 2022

[8] [8]

Yingmei Chen, Zhongyu Wei, and Xuanjing Huang. 2018. Incorporating Corpo- ration Relationship via Graph Convolutional Neural Networks for Stock Price Prediction. Proceedings of the 27th ACM International Conference on Information and Knowledge Management (2018)

work page 2018