Pith. sign in

REVIEW 4 major objections 5 minor 37 references

Joint deep calibration of the 4-factor PDV model

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read Learning SPX implied volatilities, VIX futures, and VIX call prices pointwise reduces joint calibration of the 4-factor PDV model to a few seconds of matrix-vector products.

desk verdict A credible few-second joint SPX/VIX calibration result, but the LSMC label generator is under-specified and under-validated; fix that before accept. read the letter →

arxiv 2507.09412 v1 pith:LKQXMVQO submitted 2025-07-12 q-fin.CP

Fabio Baschetti , Giacomo Bormetti , Pietro Rossi This is my paper
classification q-fin.CP MSC 65C2068T0791B7091G2091G3091G60
keywords neuralnetworksdeeppricingjointSPX/VIXcalibrationpath-dependentvolatilityleastsquaresMonteCarloVIXfuturesimpliedsurface4-factorMarkovPDVmodel
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 claims that joint calibration to SPX and VIX markets for the 4-factor Markov path-dependent volatility model can be compressed from minutes of nested Monte Carlo into about five seconds on a single CPU core. The device is to learn the pricing functions themselves: one feed-forward network maps model parameters, maturity, and strike to SPX implied volatilities, while a second network outputs VIX futures and VIX call prices together, so that every pricing call inside the calibration loop is a matrix-vector product. The inner simulation that defines the VIX is replaced offline by a ridge-regularized polynomial regression of the VIX on the four Markovian factors, which cuts the number of inner trajectories needed per training surface by a factor of about 256. If the claim holds, joint calibration becomes a near-instant routine operation for trading desks, and the offline-training recipe transfers to other models whose pricing bottleneck is nested simulation.

What carries the argument

Two feed-forward neural networks evaluated as matrix-vector products: a pointwise SPX implied-volatility network and a joint VIX network with a two-dimensional output (futures price and call price), both taking the 14-dimensional parameter vector, maturity, and strike as inputs. The offline engine that makes training feasible is a Least Squares Monte Carlo generator: the conditional expectation defining $\mathrm{VIX}_T^2$ is computed by inner simulation on a small subsample of outer paths, regressed with an L2 (ridge) penalty onto a polynomial basis in the four Markovian factors $(R_{1,0}, R_{1,1}, R_{2,0}, R_{2,1})$, and extended to the full outer sample by one matrix multiplication, cutting the inner-trajectory budget from $2^{31}$ to $2^{23}$ draws per surface (a speed-up of order 256). The joint loss function weights relative errors on SPX implied volatilities, VIX futures, and VIX implied volatilities, the last obtained by inverting the neural call price through Black's formula using the neural futures price.

What would settle it

Take a holdout set of several dozen parameter vectors spread across the full training hypercube, price VIX futures and calls by full nested Monte Carlo with $2^{18}$ outer and $2^{13}$ inner paths, and compare against the neural outputs; if a substantial share of the holdout falls outside the Monte Carlo confidence bands of Section 3, the ridge-regression labels are biasing the networks. A cheaper check is to re-run the ridge fit at parameter vectors away from the single configuration used in Figure 3 and compare the fitted polynomial against nested simulation to quantify label error directly.

Watch

Extended reading notes

Core claim

The paper's central claim is that eliminating the outer simulation from the calibration loop makes joint SPX/VIX calibration practical. The state of the art it improves on replaces only the inner simulation with a learned VIX map, leaving each loss evaluation dependent on an online outer Monte Carlo run and costing 12 to 20 minutes on a GPU. Here, two neural pricers — one for SPX implied volatilities and one for the pair (VIX futures price, VIX call price) — take model parameters and contract specifications as inputs and are evaluated on the fly. The two markets are kept consistent by a single network that outputs futures and calls together, which makes the inversion of call prices into VIX implied volatilities reliable because the neural futures price is accurate to about $7\times10^{-4}$ in 95% of test cases. The calibration loop then optimizes the ten model parameters against a weighted relative-error loss on SPX volatilities, VIX futures, and VIX implied volatilities, with all Monte Carlo cost confined to an offline training-data generation phase. Reported joint calibration time is about 5 seconds on a serial CPU.

Load-bearing premise

Every training label the networks learn from is produced by a ridge-regularized polynomial approximation of the VIX as a function of the four Markovian factors, and that approximation is validated on a single parameter configuration; the polynomial degree and ridge constant are never reported, so the label error across the full training domain is unknown.

Editorial extensions

If this is right

  • Joint calibration time falls from 12 to 20 minutes on a GPU to about 5 seconds on a serial CPU, making frequent re-calibration and intraday use feasible.
  • Any strike-maturity pair inside the training domain is priced directly, so calibration runs on market quotes without interpolation from a pre-specified grid.
  • A single network outputs VIX futures and calls together, so VIX implied volatilities derived from neural call prices inherit the futures accuracy and stay mutually consistent.
  • The Monte Carlo burden is paid once, offline, during training-data generation; afterwards pricing and calibration are instantaneous evaluations of learned maps.
  • The same pointwise-neural-pricer recipe extends to other models and to path-dependent payoffs whose pricing is bottlenecked by nested simulation.

Reading between the lines

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

  • The reported price accuracies inherit an unquantified label error, because the ridge regression that generates every VIX training label is demonstrated on one parameter set; a holdout comparison against full nested simulation would reveal whether the fitted smiles are slightly biased even where the neural fit looks exact.
  • Seconds-scale calibration changes the model-selection loop: a desk could calibrate many candidate specifications per day, or build long parameter time series across thousands of historical dates for stress testing, at a cost that was previously prohibitive.
  • The authors' observation that the learned VIX representation could replace the ridge regression inside the generator implies a possible fully neural pipeline in which the same network both labels training data and prices path-dependent exotics.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes a fully offline deep-pricing approach to joint SPX/VIX calibration in the 4-factor Markov path-dependent volatility model. Two feed-forward neural networks are trained on model-generated data: one maps model parameters and contract specifications to SPX implied volatilities, and one maps parameters, maturities, and strikes to VIX futures and call prices. The key methodological component for VIX training labels is a least-squares Monte Carlo (LSMC) procedure with ridge regression that approximates the VIX as a polynomial of the terminal factors, replacing nested simulation. After training, pricing is reduced to matrix-vector products, and the paper reports out-of-sample network errors and calibrations to two historical joint SPX/VIX surfaces, with calibration times of about 5 seconds versus 12 minutes for the hybrid neural-MC benchmark.

Significance. If the claims hold, the paper offers a substantial practical speedup for a model that is currently expensive to calibrate, and it extends the authors' earlier deep-calibration framework to the joint SPX/VIX problem. The explicit algorithms for simulation, LSMC, and neural training are useful, and the comparison of neural prices against high-precision nested Monte Carlo benchmarks on the two joint-calibration dates is a valuable sanity check. The paper also correctly emphasizes the importance of learning VIX futures accurately for the subsequent inversion to VIX implied volatilities. However, the central evidence currently rests on a label-generation step whose key hyperparameters are omitted and whose validation is limited, and on out-of-sample tests that are filtered to favorable regions. These gaps must be closed before the few-second calibration claim can be considered robust and reproducible.

major comments (4)
  1. [§3.1, Algorithm 4] The LSMC ridge regression is the sole generator of VIX training labels, but the polynomial degree d and the ridge constant c are never specified. Algorithm 4 lists them only as inputs, and the numerical validation in Figure 3 is performed on a single parameter set. Since the training domain in Section 4.3 covers a broad hypercube with no VIX smile shape filters, the accuracy of the polynomial approximation outside that single validation point is unknown. A material bias in LSMC would propagate to every neural VIX price and to the calibrated parameters in Tables 2 and 3. The paper should report d and c and provide a domain-wide comparison of LSMC against nested Monte Carlo, preferably stratified by parameter region and by futures level, showing that the bias is small relative to Monte Carlo error.
  2. [§4.3, Figures 8 and 9] The out-of-sample evaluation for VIX derivatives is conditioned on the generated futures price F falling in [0.10, 0.30], and calls are tested only when this condition holds. This selection on the target variable means the reported MAEs (2.6e-4 for futures, 7.2e-5 for calls) describe performance on a favorable subset, not on the full domain over which the network was trained. The stated rationale is that this range is what is typically found in the market, but the calibration loop is not constrained to it, and the training data itself has no such filter. The paper should report unconditional OOS errors, or at least errors stratified by F, and should verify that the calibrated parameters stay within the accurately learned region.
  3. [§4.2, Figure 6] The SPX out-of-sample test set excludes every surface that fails the shape conditions (σ_wl < 0.60 and σ_wl/σ_wr < 1.50), so the reported MAE of 7.8e-4 measures performance only on 'typical' smiles. The authors acknowledge in Section 5.1 that calibrated parameters can move outside the training domain, and the uniform buffer in the training set does not guarantee accuracy on arbitrary excluded surfaces. To support the calibration methodology, the paper should quantify the network error on the excluded surfaces, or explicitly constrain the calibration search to the validated region and report the effect of such constraints on the final fits.
  4. [§5.2, Tables 2–4] The joint calibration evidence consists of only two market dates (October 21, 2009 and April 28, 2010). Table 4 reports an average calibration time of 5 seconds, but with two dates this is not a stable basis for the paper's central speed claim. In addition, the loss function in Equation (9) contains weights w_v^SPX, w_f^VIX, and w_v^VIX whose values are never given, and the optimizer settings (algorithm, stopping criterion, initialization, number of iterations) are not detailed. The paper should include a larger calibration study with a distribution of calibration times and final losses, and should state the weights and optimizer choices needed to reproduce Table 4.
minor comments (5)
  1. [Algorithm 4] There are typos in the pseudocode: 'simulte onestep' should be 'simulate onestep', and the symbol '~V IX' appears with inconsistent spacing and tildes; these should be cleaned up.
  2. [Figure 3 caption] The caption contains 'rdge13 18x10' which should read 'ridge 18x10' or similar.
  3. [§4.1, step 5] The notation 'β1,2 = [0.05, 0.40]' is ambiguous; it should be written as an interval membership, e.g., β1,2 ∈ [0.05, 0.40], consistent with the other parameters.
  4. [Table 4] The table layout is unclear: the '# cores 1' entry under the fully neural column needs explanation, and the hardware and software environment (CPU model, MATLAB/Keras versions) used for the timing should be stated.
  5. [§5.2] The paper states that 'we train a new network to learn short SPX maturities specifically' for the joint problem, but does not describe its architecture, training data, or accuracy. Since this network is used in the joint calibration, its specifications should be reported.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the neural pricing maps are trained on model-generated data and validated against independent high-precision Monte Carlo benchmarks, so the central calibration-speed claim does not reduce to its inputs.

full rationale

The paper's claimed contribution is an empirical acceleration: replace nested Monte Carlo pricing by two feed-forward neural networks trained offline on LSMC-accelerated model prices and then calibrate model parameters by evaluating those networks. None of the calibration losses (Eq. 9) or pricing maps is defined in terms of the target parameters: the SPX implied-volatility network and the VIX futures/call network map (theta,T,K) to prices, and the training labels are generated from the 4-factor PDV dynamics, not from the market quotes being calibrated. The out-of-sample checks (Figures 6, 8, 9) and, crucially, the final fits are validated against independent high-precision nested Monte Carlo with N_out=2^18, N_inner=2^13 (Figures 10, 13, 14), so the reported few-second calibration is not a self-fulfilling fit. The LSMC ridge approximation of VIX_T (Algorithm 4) is an approximation method, not a definitional identity; the omitted polynomial degree d and ridge constant c and the validation on a single parameter set are reproducibility and accuracy concerns, not circularity. Self-citations to Baschetti et al. (2024) provide the previously published SPX deep-pricing methodology and are not load-bearing for the new joint VIX result; moreover the SPX network accuracy is re-measured here (Figure 6). The training-domain bounds informed by 50 calibrated SPX surfaces are a data-driven prior, not a fitted output renamed as a prediction. No equation in the paper reduces a predicted quantity to its own input by construction.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the accuracy of LSMC-generated training labels and on the neural networks' generalization within and slightly outside a hand-picked training domain. The paper reports strong out-of-sample numbers but only on filtered test sets, and it leaves the polynomial degree and ridge constant unspecified.

free parameters (5)
  • polynomial degree d in LSMC = not stated
    Controls the monomial basis for VIX regression (Algorithm 4, Section 3.1); the paper never specifies d, and the accuracy of all training labels depends on it.
  • ridge regularization constant c = not stated
    L2 penalty in the LSMC regression (Section 3.1, step iii); stability of coefficients relies on c, but its value is omitted.
  • volatility cap for simulation = 1.5
    Caps sigma_t at 1.5 in Algorithm 1 following Guyon and Lekeufack (2023); the cap changes the dynamics of R1 and R2 and is 'quite arbitrary' (Section 2.3).
  • training parameter bounds = intervals in Section 4.1
    Bounds for beta's, lambda's, and factor initial values are chosen by hand and by calibrating 50 SPX surfaces; they determine the network's domain and hence its extrapolation behavior.
  • test-set filters = shape constraints; F in [0.10, 0.30]
    Out-of-sample errors are computed only on filtered subsets, so the reported accuracy applies to a restricted region of parameter space.
assumptions (4)
  • domain assumption The 4-factor Markov PDV model with the parabolic volatility function (Equation 2) is an adequate pricing model for SPX and VIX derivatives.
    The paper takes the model from Guyon and Lekeufack (2023) as given and builds a calibrator for it; it does not test the model's adequacy beyond noting limitations for long maturities.
  • standard math The underlying S has no jumps, so the model VIX_T^2 equals (1/Delta) times the integral from T to T+Delta of E[sigma_u^2 | F_T] du (Equation 7).
    Used for pricing VIX derivatives; a standard Ito argument. The market VIX is defined by option prices (Equation 5), and the paper assumes equivalence without modeling jumps.
  • domain assumption LSMC with a monomial basis and ridge penalty accurately approximates VIX_T across the training domain.
    Section 3.1 demonstrates the polynomial fit on one parameter set (Figure 3) and uses it to generate all training labels; degree d and ridge c are unspecified, so the error across the domain is not quantified.
  • domain assumption Neural networks trained on the sampled domain generalize to real market calibration points, including points outside the training bounds.
    Section 5.1 reports a July 13, 2016 case where R1,0,0 exceeds its upper training bound and the network still extrapolates well; the authors state there is no theoretical guarantee of extrapolation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Joint deep calibration of the 4-factor PDV model." pith.science (2026). https://pith.science/paper/LKQXMVQO

@misc{pith2026250709412,
  author       = {Pith},
  title        = {Pith review of: Joint deep calibration of the 4-factor PDV model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/LKQXMVQO}},
  note         = {Machine review of arXiv:2507.09412}
}
read the original abstract

Joint calibration to SPX and VIX market data is a delicate task that requires sophisticated modeling and incurs significant computational costs. The latter is especially true when pricing of volatility derivatives hinges on nested Monte Carlo simulation. One such example is the 4-factor Markov Path-Dependent Volatility (PDV) model of Guyon and Lekeufack (2023). Nonetheless, its realism has earned it considerable attention in recent years. Gazzani and Guyon (2025) marked a relevant contribution by learning the VIX as a random variable, i.e., a measurable function of the model parameters and the Markovian factors. A neural network replaces the inner simulation and makes the joint calibration problem accessible. However, the minimization loop remains slow due to expensive outer simulation. The present paper overcomes this limitation by learning SPX implied volatilities, VIX futures, and VIX call option prices. The pricing functions reduce to simple matrix-vector products that can be evaluated on the fly, shrinking calibration times to just a few seconds.

Figures

Figures reproduced from arXiv: 2507.09412 by the authors.

Figure 1
Figure 1. 4-factor Markov PDV model smiles and terminal-price histo [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Convergence of V IX2 T estimates and its effect on the smile. LHS: behavior of the Monte Carlo estimate of V IX2 T as a function of the number of points included in the mean. RHS: comparison between the benchmark smile (green) and the prices obtained when reducing to log2 Nsim inn = 10 (blue). Dashed black lines represent 95% confidence bands. log2 Nsim out = 18 and the number of time steps per day is six. Model par… view at source ↗
Figure 3
Figure 3. Quality of the ridge regression for V IXT and its effect on the smile. LHS: V IXT as computed by inner simulation vs its polynomial approximation V IX [T . RHS: comparison between the benchmark smile (green) and the prices obtained by Least Squares Monte Carlo with log2 Nsub = 13 and log2 Nsim inn = 10 (blue). Black lines correspond to 95% confidence bands. log2 Nsim out = 18 and the number of time steps per day is … view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Neural network candidate for the futures price as a fun [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 5
Figure 5. Figure 5: Histograms of the model parameters as sampled for the t [PITH_FULL_IMAGE:figures/full_fig_p016_5.png]
Figure 6
Figure 6. Figure 6: Out-of-sample performance of the neural approximatio [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 7
Figure 7. Figure 7: Neural network architecture for VIX derivatives. The in [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: Out-of-sample performance of the neural approximatio [PITH_FULL_IMAGE:figures/full_fig_p019_8.png]
Figure 9
Figure 9. Figure 9: Out-of-sample error distribution of the VIX futures (1- [PITH_FULL_IMAGE:figures/full_fig_p019_9.png]
Figure 10
Figure 10. Figure 10: SPX calibration of the 4-factor Markov PDV model as of J [PITH_FULL_IMAGE:figures/full_fig_p022_10.png]
Figure 11
Figure 11. Figure 11: MSE (left) and MAE (right) over 43 SPX volatility surfaces [PITH_FULL_IMAGE:figures/full_fig_p023_11.png]
Figure 12
Figure 12. Figure 12: MSE (left) and MAE (right) over 43 SPX volatility surfaces [PITH_FULL_IMAGE:figures/full_fig_p023_12.png]
Figure 13
Figure 13. Figure 13: Joint SPX-VIX calibration of the 4-factor Markov PDV mo [PITH_FULL_IMAGE:figures/full_fig_p024_13.png]
Figure 14
Figure 14. Figure 14: Joint SPX-VIX calibration of the 4-factor Markov PDV mo [PITH_FULL_IMAGE:figures/full_fig_p025_14.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

37 extracted references · 37 canonical work pages

  1. [1]

    Abi Jaber, C

    E. Abi Jaber, C. Illand, and L. S. Joint SPX & VIX calibration with G aussian polynomial volatility models: Deep pricing with quantization hints. Mathematical Finance, 2025

    work page 2025

  2. [2]

    Andr\'es, A

    H. Andr\'es, A. Boumezoued, and B. Jourdain. Implied volatility (also) is path-dependent. Preprint arXiv:2312.15950, 2023

    work page arXiv 2023

  3. [3]

    Baschetti, G

    F. Baschetti, G. Bormetti, S. Romagnoli, and P. Rossi. The SINC way: A fast and accurate approach to F ourier pricing. Quantitative Finance, 22 0 (3): 0 427--446, 2022

    work page 2022

  4. [4]

    Baschetti, G

    F. Baschetti, G. Bormetti, and P. Rossi. Deep calibration with random grids. Quantitative Finance, 24 0 (9): 0 1263--1285, 2024

    work page 2024

  5. [5]

    Bayer and B

    C. Bayer and B. Stemper. Deep calibration of rough stochastic volatility models. Working paper, 2018

    work page 2018

  6. [6]

    Bayer, P

    C. Bayer, P. Friz, and J. Gatheral. Pricing under rough volatility. Quantitative Finance, 16 0 (6): 0 887--904, 2016

    work page 2016

  7. [7]

    L. Bergomi. Smile dynamics II . Available at SSRN 1493302, 2005

    work page 2005

  8. [8]

    Blanc, B

    P. Blanc, B. Donier, and B. J.P. Quadratic H awkes processes for financial prices. Quantitative Finance, 17 0 (2): 0 171--188, 2017

    work page 2017

Show all 37 references
  1. [9]

    Bondi, S

    A. Bondi, S. Pulido, and S. Scotti. The rough H awkes H eston stochastic volatility model. Mathematical Finance, 0 (34): 0 1197--1241, 2024

    2024

  2. [10]

    Bourgey and J

    F. Bourgey and J. Guyon. Fast exact joint S & P 500/ VIX smile calibration in discrete and continuous time. Risk, 2024

    2024

  3. [11]

    Chicheportiche and J

    R. Chicheportiche and J. Bouchaud. The fine-structure of volatility feedback I : Multi-scale self-reflexivity. Physica A: Statistical Mechanics and its Applications, 410: 0 174--195, 2014

    2014

  4. [12]

    Cuchiero, M

    C. Cuchiero, M. Keller-Ressel, and J. Teichmann. Polynomial processes and their applications to mathematical finance. Finance and Stochastics, 16: 0 711--740, 2012

    2012

  5. [13]

    Cuchiero, G

    C. Cuchiero, G. Gazzani, and S. Svaluto-Ferro. Signature-based models: Theory and calibration. SIAM Journal on Financial Mathematics, 14 0 (3): 0 910--957, 2022

    2022

  6. [14]

    Cuchiero, G

    C. Cuchiero, G. Gazzani, J. M\"oller, and S. Svaluto-Ferro. Joint calibration to SPX and VIX options with signature-based model. Mathematical Finance, pages 1--54, 2024

    2024

  7. [15]

    Filipovi \'c and M

    D. Filipovi \'c and M. Larsson. Polynomial diffusions and applications in finance. Finance and Stochastics, 20 0 (4): 0 931--972, 2016

    2016

  8. [16]

    Foschi and A

    P. Foschi and A. Pascucci. Calibration of the H obson and R ogers model: empirical tests. Computational Statiscs & Data Analysis, 53: 0 2219–--2235, 2005

    2005

  9. [17]

    Gatheral, T

    J. Gatheral, T. Jaisson, and M. Rosenbaum. Volatility is rough. Quantitative Finance, 18 0 (6): 0 933--949, 2018

    2018

  10. [18]

    Gatheral, P

    J. Gatheral, P. Jusselin, and M. Rosenbaum. The quadratic rough H eston model and the joint calibration problem. Risk, 2020

    2020

  11. [19]

    Gazzani and J

    G. Gazzani and J. Guyon. Pricing and calibration in the 4-factor path-dependent volatility model. Quantitative Finance, 25 0 (3): 0 471--489, 2025

    2025

  12. [20]

    Guerreiro and J

    H. Guerreiro and J. Guerra. Least squares M onte C arlo methods in stochastic V olterra rough volatility models. preprint arXiv:2105.04511, 2021

    2021 arXiv

  13. [21]

    I. Guo, G. Loeper, J. Obloj, and S. Wang. Optimal transport for model calibration. Risk, 2022

    2022

  14. [22]

    J. Guyon. Path-dependent volatility. Risk, 2014

    2014

  15. [23]

    J. Guyon. Dispersion-constrained martingale S chr\"odinger problems and the exact joint S & P 500/ VIX smile calibration puzzle. Finance and Stochastics, 28 0 (1): 0 27--79, 2024

    2024

  16. [24]

    Guyon and J

    J. Guyon and J. Lekeufack. Volatility is (mostly) path-dependent. Quantitative Finance, 23 0 (9): 0 1221--1258, 2023

    2023

  17. [25]

    Guyon and S

    J. Guyon and S. Mustapha. Neural joint S&P 500/ VIX smile calibration. Available at SSRN, 2022

    2022

  18. [26]

    Hobson and L

    D. Hobson and L. Rogers. Complete models with stochastic volatility. Mathematical Finance, 8 0 (1): 0 11--27, 1998

    1998

  19. [27]

    Horvath, A

    B. Horvath, A. Muguruza, and M. Tomas. Deep learning volatility: a deep neural network perspective on pricing and calibration in (rough) volatility models. Quantitative Finance, 21 0 (1): 0 11--27, 2021

    2021

  20. [28]

    A. Lewis. Option valuation under stochastic volatility with M athematica code . Finance Press, 2000

    2000

  21. [29]

    A. L. Lewis. A simple option formula for general jump-diffusion and other exponential L \'e vy processes. Available at SSRN 282110, 2001

    2001

  22. [30]

    S. Liu, A. Borovykh, L. Grzelak, and C. Oosterlee. A neural network-based framework for financial model calibration. Journal of Mathematics in Industry, 9 0 (1): 0 1--28, 2019

    2019

  23. [31]

    Pacati, G

    C. Pacati, G. Pompa, and R. Ren \`o . Smiling twice: the H eston++ model. Journal of Banking & Finance, 96: 0 185--206, 2018

    2018

  24. [32]

    L. Parent. The EWMA H eston model. Quantitative Finance, 23 0 (1): 0 71--93, 2023

    2023

  25. [33]

    S. R mer. Empirical analysis of rough and classical stochastic volatility models to the SPX and VIX markets. Quantitative Finance, 22 0 (10): 0 1805--1838, 2022

    2022

  26. [34]

    Rosenbaum and J

    M. Rosenbaum and J. Zhang. Deep calibration of the quadratic rough H eston model. preprint arXiv:2107.01611, 2021

    2021 arXiv

  27. [35]

    E. Sentana. Quadratic ARCH models. The Review of Economic Studies, 62 0 (4): 0 639--661, 1995

    1995

  28. [36]

    G. Zumbach. Time reversal invariance in finance. Quantitative Finance, 9 0 (5): 0 505--515, 2009

    2009

  29. [37]

    G. Zumbach. Volatility conditional on price trends. Quantitative Finance, 10 0 (4): 0 431--442, 2010

    2010

Pith tools

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