Pith. sign in

REVIEW 5 major objections 3 minor 44 references

Neural Jumps for Option Pricing

T0 review · 5 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a neural jump-diffusion model, trained by backpropagation through a Gumbel-Softmax relaxation, produces the most accurate S&P 500 option prices among six competing models.

desk verdict A plausible neural jump-SDE option pricing model that beats benchmarks in MSE, but the jump mechanism is under-specified to the point where the attribution to 'neural jumps' is not supported. read the letter →

arxiv 2506.05137 v1 pith:OLOMEKVT submitted 2025-06-05 q-fin.GN

Duosi Zheng , Hanzhong Guo , Yanchu Liu , Wei Huang This is my paper
classification q-fin.GN MSC 91G2068T0791G60
keywords optionpricingjumpdiffusionneuralstochasticdifferentialequationsGumbel-SoftmaxS&P500indexoptionsdeeplearningSVCJmodelMonteCarlosimulation
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 argues that jump risk can be brought into deep-learning-based option pricing by building a neural jump stochastic differential equation (NJSDE) model: keep the stochastic structure of a jump diffusion, but let neural networks estimate the drift, diffusion, jump intensity, and jump-size coefficients. The central technical move is to make the discrete jump process differentiable with a Gumbel-Softmax relaxation, so the whole model can be trained by backpropagation through Monte Carlo price paths. On simulated data from a stochastic-volatility-with-correlated-jumps model, the NJSDE model has the lowest pricing errors among all benchmarks, while on simulated Heston data it matches the jump-free neural SDE model in out-of-sample accuracy. On 591,284 S&P 500 call options, the NJSDE model records the lowest out-of-sample MAE and MSE of the six compared models, with a 32 percent lower MSE than its closest neural competitor. A sympathetic reader would take the paper as establishing that hybrid structural-neural models need not ignore jumps and that such models can be optimized with ordinary gradient methods.

What carries the argument

The mechanism that carries the argument is the discretized NJSDE recursion, which is treated as a recurrent neural network so that the loss function's gradient can be computed by the chain rule and backpropagation. The jump term is written as the product $NN_3U_t^{(S)}f(NN_7)$ for the asset price and $NN_6U_t^{(V)}f(NN_7)$ for volatility, where $U_t$ is a uniform-(0,1) random variable that reparameterizes the jump size and $f(NN_7)$ is the Gumbel-Softmax relaxation of the number of jumps with intensity estimated by $NN_7$. Eight independent networks share the inputs $(S_t,K,\tau,r_f)$ and together replace the drift, diffusion, jump, and correlation parameters of the SVCJ model, with the correlation encoded by $NN_8$ through $dW_t^{(V)}=NN_8dW_t^{(S)}+\sqrt{1-NN_8^2}dW_t$. Monte Carlo paths give price estimates, and because every component in the discretized recursion is differentiable, the gradient of the squared pricing error with respect to each network parameter can be propagated backward through time.

What would settle it

Generate synthetic option prices from an SVCJ model whose jump sizes have a negative mean, then fit the NJSDE model and compare out-of-sample MSE against the correctly specified SVCJ model. Since the NJSDE's uniform-(0,1) reparameterization can only produce positive jumps, its out-of-sample error should be visibly worse than the parametric benchmark if jumps are what drive the paper's reported advantage.

Watch

Extended reading notes

Core claim

The core discovery is that a jump-diffusion model whose deterministic coefficients are replaced by feedforward neural networks can be trained end-to-end with backpropagation, and that the resulting model prices options more accurately than parametric jump models, nonparametric networks, and a jump-free neural SDE model. The paper starts from the SVCJ model, writes the discretized dynamics as $S_{t+\Delta t}=S_t+NN_1\Delta t+NN_2\sqrt{\Delta t}\varepsilon_t^{(S)}+NN_3 U_t^{(S)}f(NN_7)$ and an analogous equation for volatility, and uses the Gumbel-Softmax approximation $f(NN_7)$ to relax the Poisson jump-counting process. In the empirical exercise the model's out-of-sample MSE on S&P 500 options is 10.64, compared with 15.63 for the jump-free neural SDE and 46.35 for the parametric SVCJ model, which the authors read as evidence that jumps are present in the option surface and are captured by the neural jump component.

Load-bearing premise

The load-bearing premise is that the true jump process is adequately represented by a Gumbel-Softmax relaxation of the jump count and by jump sizes of the form $NN_3 U_t$ with $U_t$ uniform on $(0,1)$; if real jumps are negative, heavy-tailed, or correlated with volatility, this representation is misspecified and the accuracy gains cannot be attributed to a correct jump model.

Editorial extensions

If this is right

  • The backpropagation-compatibility result implies that jump processes can be included in neural SDE option-pricing models without switching to adjoint sensitivity or likelihood-based training methods.
  • On large panel data, the jump-aware hybrid outperforms both the parametric SVCJ model and the ANN model out of sample, suggesting that structural constraints plus jumps dominate either ingredient alone.
  • In a setting without genuine jumps, the learned jump coefficients shrink and the NJSDE model behaves like the jump-free NSDE model, so the extra component does not destroy accuracy when it is unnecessary.
  • The significant DM test statistics against the NSDE model on S&P 500 options imply that the accuracy gain is not just a small-sample artifact, and that jump terms carry real pricing information even after neural flexibility is allowed.
  • Because the model retains stochastic structure, its price estimates remain economically interpretable as present values of Monte Carlo payoffs, in contrast to purely black-box ANN pricing surfaces.

Reading between the lines

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

  • Editorial extension: The paper models jump sizes as $NN_3 U_t$ with $U_t\sim \text{Uniform}(0,1)$, which forces all jumps positive; a natural stress test is to generate SVCJ data with negative-mean jumps, where this reparameterization cannot match the true dynamics and the reported advantage should shrink or reverse.
  • Editorial extension: The S&P 500 sample spans 2018-2022 and includes the March 2020 crash, so part of the jump component's gain may come from absorbing crisis-period outliers; testing on a calm subsample would separate crisis handling from steady-state jump risk.
  • Editorial extension: The paper leaves the temperature $\tau$ and truncation bound $n$ of the Gumbel-Softmax approximation as unspecified hyperparameters; reporting sensitivity to them would show how much of the empirical gain depends on tuning the relaxation.
  • Editorial extension: One could recover the fitted $NN_7$ and $NN_3$ outputs at each date to infer an implied jump-intensity surface, connecting the pricing improvement to economic jump-risk premia rather than only price errors.
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

5 major / 3 minor

Summary. The paper proposes a Neural Jump Stochastic Differential Equation (NJSDE) model for option pricing, in which deterministic coefficients of an SVCJ-type jump-diffusion model are replaced by neural networks and the jump-count mechanism is made differentiable through Gumbel-Softmax relaxation. The model is calibrated by Monte Carlo simulation and backpropagation, and it is tested on synthetic Heston and SVCJ data as well as on 591,284 S&P 500 index options. The reported in-sample and out-of-sample MAE/MSE metrics and Diebold-Mariano tests indicate that NJSDE often has the lowest pricing errors, with the largest out-of-sample gains over NSDE in the real-data experiment.

Significance. The empirical setup is ambitious: it includes a large OptionMetrics sample, several parametric and nonparametric benchmarks, and Diebold-Mariano significance tests. If the model were correctly specified and reproducible, an out-of-sample MSE improvement from 15.63 (NSDE) to 10.64 (NJSDE) on S&P 500 options would be a meaningful result for the neural option-pricing literature. However, the core model is incompletely specified: the function f(NN7) is never defined, the jump-size reparameterization cannot represent the SVCJ jump law, and the discretization is inconsistent with the stated log-price SDE. These issues prevent me from assessing whether the reported accuracy gains reflect a correctly specified neural jump process or merely added model flexibility. No code, architecture details, or key hyperparameters are provided, so the empirical claims cannot be verified from the manuscript as written.

major comments (5)
  1. [Section 2.2, Eqs. (6)-(8), (11)-(12)] The function f(NN7) is never defined. The Gumbel-Softmax construction in Eqs. (1)-(3) yields relaxed probabilities y_i, but the paper never states how these become the scalar f(NN7) that multiplies the jump term in Eq. (6), nor how the jump count enters the Monte Carlo path. Eq. (8) only defines π_i. As a result, Equations (11)-(12), the gradient formula in Eq. (17), and Algorithm 1 cannot be implemented or checked. Please define f explicitly, state its inputs and output, and show that it is differentiable with respect to NN7.
  2. [Section 2.2, Eq. (6); Section 3.2, Table 3] The jump-size specification NN3 U_t^(S) f(NN7) cannot represent the SVCJ jump law used in Section 3.2. In SVCJ, Z^y_t is Gaussian, unbounded, two-sided, and correlated with Z^v_t; by contrast, U_t^(S) is uniform on (0,1) and NN3 is a deterministic function of (S_t, K, t, r_f), so the jump has fixed sign and bounded magnitude and cannot depend on a per-jump volatility shock. Consequently Tables 3 and 4 do not test the model's ability to capture SVCJ jump dynamics, and the observed improvement over benchmarks could be due to unconstrained flexibility rather than a correctly specified jump process. Please replace the uniform reparameterization with a distribution that matches SVCJ, or clearly justify a different target model, and re-run the experiments.
  3. [Section 2.3, Eq. (11); Section 2.2, Eq. (4)] The discretized model is not a valid Euler approximation of the stated SDE. Eq. (4) is a log-price SDE, while Eq. (11) is an arithmetic update for the price level with no exponential transformation or Itô correction, so the discretized dynamics do not correspond to the SVCJ process claimed in Section 2.2. In addition, because all NNi take K and t as inputs, the SDE coefficients are option-specific, meaning that each strike generates its own underlying price process. This breaks the structural interpretation in Section 1 and may violate no-arbitrage. Please derive a consistent discretization and clarify whether the model is intended as an asset-price process or as a pricing-surface regression.
  4. [Section 2.2, Eq. (9)] Equation (9) uses sqrt(1 - (NN8)^2), but no constraint is placed on NN8. If NN8 is an arbitrary neural network output, it can leave the interval [-1,1] during training, making the volatility correlation ill-defined. Please parameterize NN8 with a bounded activation such as tanh, or otherwise constrain it, and report the learned correlation values.
  5. [Section 3, Algorithm 1; Section 4] The empirical results are not reproducible as reported. The paper does not specify the network architectures (number of layers, units, activation functions), the optimizer and learning rate, the number of epochs D, the number of Monte Carlo paths M, the number of time steps m, the Gumbel-Softmax temperature τ, or the truncation bound n. Because the central claim of the paper is an empirical accuracy gain, these implementation details are load-bearing; please report them completely and, ideally, release the code used to produce Tables 1, 3, and 6.
minor comments (3)
  1. [Section 1] The sentence 'the ANN method does not have a sounding theory' should read 'sound theory'.
  2. [Section 3.1] The text states that the NJSDE model can degenerate into the NSDE model when jump coefficients converge to zero, but no learned NN3/NN7 values, shrinkage mechanism, or formal nesting argument is reported; please verify this claim with evidence.
  3. [Tables 2, 4, and 7] Please state which loss function (MAE or MSE) is used in the Diebold-Mariano tests, and add standard errors or confidence intervals for the MSE estimates in Tables 1, 3, and 6 so that the magnitudes of the reported gains are interpretable.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the NJSDE pricing comparisons are genuine out-of-sample evaluations against independent benchmarks; the undefined jump function is a specification concern, not a circular step.

full rationale

The paper's central claim is that a neural jump SDE improves option-pricing accuracy. The chain of evidence is empirical: the model is calibrated by minimizing squared pricing error (Eq. 13) on training options and then evaluated on held-out strikes and maturities, both in simulated experiments (Tables 1 and 3) and on S&P 500 data (Table 6). The target prices are external to the model or generated from independent parametric processes; the NJSDE's out-of-sample prices are not constructed from those targets, so the 'prediction' is not a fitted input renamed as a forecast. The benchmarks (BS, Heston, SVCJ, ANN, NSDE) are independent models, not outputs of the NJSDE. There is no load-bearing self-citation: Wang and Hong (2021) appears as the NSDE benchmark and as a methodological antecedent, but the NJSDE result is not derived from that reference, and none of the cited prior work shares the present authors. The only notable weakness is that f(NN7) is never explicitly defined in Section 2.2, making the jump term's exact distribution ambiguous; this is a model-identification and correctness risk, not a circularity, because it does not make the model's output equal to its training target by construction. The claimed contribution therefore stands or falls on out-of-sample accuracy, which is a non-circular form of evidence.

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

The central empirical claim rests on trained neural network weights and on several unreported hyperparameters. The modeling assumptions above are the most consequential: the Gumbel-Softmax relaxation, the uniform jump-size representation, and the arithmetic discretization are all load-bearing and none is validated.

free parameters (4)
  • Gumbel-Softmax temperature tau
    Controls how close the relaxed jump samples are to one-hot; neither its value nor a schedule is reported, yet it determines gradient bias in the jump term.
  • Maximum number of jumps per step n
    User-specified truncation in Eq. (1) that defines the categorical support; not reported.
  • Simulation hyperparameters M and m
    Number of Monte Carlo paths and Euler time steps per maturity in Algorithm 1; not reported.
  • Network architecture and optimizer hyperparameters
    Depth, width, activations, learning rate, epochs, and batch size for the eight networks NNi are not given, so the model cannot be re-implemented from the paper.
assumptions (5)
  • standard math Standard Ito calculus and Euler-Maruyama discretization converge for the SDE system in Eqs. (11)-(12)
    The paper discretizes dSt and dVt with Euler steps and treats the recursion as a recurrent network.
  • domain assumption Gumbel-Softmax samples provide unbiased or sufficiently low-bias gradients for the jump parameters
    Invoked via Jang et al. (2017) in Section 2.1, but temperature tau is not reported and bias is not analyzed.
  • standard math A neural network can approximate the coefficient functions of the SVCJ model
    Invoked via Hornik et al. (1989) in Section 1; universal approximation is a theorem, but the required network size is not addressed.
  • ad hoc to paper Uniform(0,1) reparameterization can represent the jump size distributions of SVCJ
    Section 2.2 replaces Z^y and Z^v, which in SVCJ are correlated and can be negative or heavy-tailed, with U~(0,1) times NN outputs; this is not justified and limits the jump model.
  • ad hoc to paper The additive Euler update St+dt = St + ... is a valid approximation of the log-price SVCJ dynamics in Eq. (4)
    The original SVCJ is written for d log St; the discretized model updates St arithmetically, changing the distributional properties of returns.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Neural Jumps for Option Pricing." pith.science (2026). https://pith.science/paper/OLOMEKVT

@misc{pith2026250605137,
  author       = {Pith},
  title        = {Pith review of: Neural Jumps for Option Pricing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OLOMEKVT}},
  note         = {Machine review of arXiv:2506.05137}
}
read the original abstract

Recognizing the importance of jump risk in option pricing, we propose a neural jump stochastic differential equation model in this paper, which integrates neural networks as parameter estimators in the conventional jump diffusion model. To overcome the problem that the backpropagation algorithm is not compatible with the jump process, we use the Gumbel-Softmax method to make the jump parameter gradient learnable. We examine the proposed model using both simulated data and S&P 500 index options. The findings demonstrate that the incorporation of neural jump components substantially improves the accuracy of pricing compared to existing benchmark models.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

44 extracted references · 42 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

    work page

  2. [2]

    , " * write output.state after.block = add.period write newline

    ENTRY address author booktitle chapter doi edition editor eid howpublished institution isbn issn journal key month note number organization pages publisher school series title type url volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.all := #1...

  3. [3]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in "" FUNCTION format.date year ...

    work page

  4. [4]

    , year 2003

    author Amilon, H. , year 2003 . title A neural network versus Black--Scholes : A comparison of pricing and hedging performances . journal Journal of Forecasting volume 22 , pages 317--335

    work page 2003

  5. [5]

    , author Charalambous, C

    author Andreou, P.C. , author Charalambous, C. , author Martzoukos, S.H. , year 2008 . title Pricing and trading European options by combining artificial neural networks and parametric models with implied parameters . journal European Journal of Operational Research volume 185 , pages 1415--1433

    work page 2008

  6. [6]

    , author Charalambous, C

    author Andreou, P.C. , author Charalambous, C. , author Martzoukos, S.H. , year 2010 . title Generalized parameter functions for option pricing . journal Journal of Banking & Finance volume 34 , pages 633--646

    work page 2010

  7. [7]

    , year 1996

    author Bates, D.S. , year 1996 . title Jumps and stochastic volatility: Exchange rate processes implicit in deutsche mark options . journal The Review of Financial Studies volume 9 , pages 69--107

    work page 1996

  8. [8]

    , year 2003

    author Bates, D.S. , year 2003 . title Empirical option pricing: A retrospection . journal Journal of Econometrics volume 116 , pages 387--404

    work page 2003

Show all 44 references
  1. [9]

    , author Scholes, M

    author Black, F. , author Scholes, M. , year 1973 . title The pricing of options and corporate liabilities . journal Journal of Political Economy volume 81 , pages 637--654

    1973

  2. [10]

    , author Liu, X

    author Cao, Y. , author Liu, X. , author Zhai, J. , year 2021 . title Option valuation under no-arbitrage constraints with neural networks . journal European Journal of Operational Research volume 293 , pages 361--374

    2021

  3. [11]

    , author Didisheim, A

    author Chen, H. , author Didisheim, A. , author Scheidegger, S. , year 2025 . title Deep surrogates for finance: With an application to option pricing . journal Journal of Financial Economics. note Forthcoming

    2025

  4. [12]

    , author Rubanova, Y

    author Chen, R.T. , author Rubanova, Y. , author Bettencourt, J. , author Duvenaud, D.K. , year 2018 . title Neural ordinary differential equations . journal Advances in Neural Information Processing Systems (NeurIPS) volume 31

    2018

  5. [13]

    , author Tsai, W.C

    author Chung, S.L. , author Tsai, W.C. , author Wang, Y.H. , author Weng, P.S. , year 2011 . title The information content of the S&P 500 index and VIX options on the dynamics of the S&P 500 index . journal Journal of Futures Markets volume 31 , pages 1170--1201

    2011

  6. [14]

    , author Ross, S.A

    author Cox, J.C. , author Ross, S.A. , year 1976 . title The valuation of options for alternative stochastic processes . journal Journal of Financial Economics volume 3 , pages 145--166

    1976

  7. [15]

    , author Esposito, F

    author Cummins, M. , author Esposito, F. , year 2025 . title Appraising model complexity in option pricing . journal Journal of Futures Markets volume 45 , pages 455--472

    2025

  8. [16]

    , author Padhy, S

    author Das, S.P. , author Padhy, S. , year 2017 . title A new hybrid parametric and machine learning model with homogeneity hint for European-style index option pricing . journal Neural Computing and Applications volume 28 , pages 4061--4077

    2017

  9. [17]

    , author Mariano, R.S

    author Diebold, F.X. , author Mariano, R.S. , year 2002 . title Comparing predictive accuracy . journal Journal of Business & Economic Statistics volume 20 , pages 134--144

    2002

  10. [18]

    , author Pan, J

    author Duffie, D. , author Pan, J. , author Singleton, K. , year 2000 . title Transform analysis and asset pricing for affine jump-diffusions . journal Econometrica volume 68 , pages 1343--1376

    2000

  11. [19]

    , et al., year 1994

    author Dupire, B. , et al., year 1994 . title Pricing with a smile . journal Risk volume 7 , pages 18--20

    1994

  12. [20]

    , year 2017

    author E, W. , year 2017 . title A proposal on machine learning via dynamical systems . journal Communications in Mathematics and Statistics volume 1 , pages 1--11

    2017

  13. [21]

    , author Johannes, M

    author Eraker, B. , author Johannes, M. , author Polson, N. , year 2003 . title The impact of jumps in volatility and returns . journal The Journal of Finance volume 58 , pages 1269--1300

    2003

  14. [22]

    , year 1954

    author Gumbel, E.J. , year 1954 . title Statistical theory of extreme values and some practical applications: A series of lectures . publisher US Government Printing Office

    1954

  15. [23]

    , year 2023

    author Halskov, K. , year 2023 . title A deep structural model for empirical asset pricing . note Working Paper

    2023

  16. [24]

    , year 1993

    author Heston, S.L. , year 1993 . title A closed-form solution for options with stochastic volatility with applications to bond and currency options . journal The Review of Financial Studies volume 6 , pages 327--343

    1993

  17. [25]

    , author Stinchcombe, M

    author Hornik, K. , author Stinchcombe, M. , author White, H. , year 1989 . title Multilayer feedforward networks are universal approximators . journal Neural Networks volume 2 , pages 359--366

    1989

  18. [26]

    , author White, A

    author Hull, J. , author White, A. , year 1987 . title The pricing of options on assets with stochastic volatilities . journal The Journal of Finance volume 42 , pages 281--300

    1987

  19. [27]

    , author Lo, A.W

    author Hutchinson, J.M. , author Lo, A.W. , author Poggio, T. , year 1994 . title A nonparametric approach to pricing and hedging derivative securities via learning networks . journal The Journal of Finance volume 49 , pages 851--889

    1994

  20. [28]

    , author Gu, S

    author Jang, E. , author Gu, S. , author Poole, B. , year 2017 . title Categorical reparameterization with Gumbel-Softmax . journal International Conference on Learning Representations (ICLR)

    2017

  21. [29]

    , author Benson, A.R

    author Jia, J. , author Benson, A.R. , year 2019 . title Neural jump stochastic differential equations . journal Advances in Neural Information Processing Systems (NeurIPS) volume 32

    2019

  22. [30]

    , author Lu, J

    author Khoo, Y. , author Lu, J. , author Ying, L. , year 2021 . title Solving parametric PDE problems with artificial neural networks . journal European Journal of Applied Mathematics volume 32 , pages 421--435

    2021

  23. [31]

    , author Morrill, J

    author Kidger, P. , author Morrill, J. , author Foster, J. , author Lyons, T. , year 2020 . title Neural controlled differential equations for irregular time series . journal Advances in Neural Information Processing Systems (NeurIPS) volume 33 , pages 6696--6707

    2020

  24. [32]

    , year 2021

    author Kim, S. , year 2021 . title Portfolio of volatility smiles versus volatility surface: Implications for pricing and hedging options . journal Journal of Futures Markets volume 41 , pages 1154--1176

    2021

  25. [33]

    , year 2002

    author Kou, S.G. , year 2002 . title A jump-diffusion model for option pricing . journal Management Science volume 48 , pages 1086--1101

    2002

  26. [34]

    , author Wong, T.K.L

    author Li, X. , author Wong, T.K.L. , author Chen, R.T. , author Duvenaud, D. , year 2020 . title Scalable gradients for stochastic differential equations . journal International Conference on Artificial Intelligence and Statistics (AISTATS) , pages 3870--3882

    2020

  27. [35]

    , author Oosterlee, C.W

    author Liu, S. , author Oosterlee, C.W. , author Bohte, S.M. , year 2019 . title Pricing options and computing implied volatilities using neural networks . journal Risk volume 7 , pages 16

    2019

  28. [36]

    , year 2015

    author Ludwig, M. , year 2015 . title Robust estimation of shape-constrained state price density surfaces . journal The Journal of Derivatives volume 22 , pages 56--72

    2015

  29. [37]

    , author Wu, X

    author Ma, J. , author Wu, X. , author Wu, H. , year 2023 . title A neural network rough volatility model . note Working Paper

    2023

  30. [38]

    , author Dixit, V

    author Ma, Y. , author Dixit, V. , author Innes, M.J. , author Guo, X. , author Rackauckas, C. , year 2021 . title A comparison of automatic differentiation and continuous sensitivity analysis for derivatives of differential equation solutions . journal 2021 IEEE High Performa...

    2021

  31. [39]

    , year 1976

    author Merton, R.C. , year 1976 . title Option pricing when underlying stock returns are discontinuous . journal Journal of Financial Economics volume 3 , pages 125--144

    1976

  32. [40]

    , author Wang, W

    author Ruf, J. , author Wang, W. , year 2020 . title Neural networks for option pricing and hedging: A literature review . journal Journal of Computational Finance volume 24 , pages 1--46

    2020

  33. [41]

    , author Wang, W

    author Ruf, J. , author Wang, W. , year 2022 . title Hedging with linear regressions and neural networks . journal Journal of Business & Economic Statistics volume 40 , pages 1442--1454

    2022

  34. [42]

    , author Zhu, S.P

    author Shvimer, Y. , author Zhu, S.P. , year 2024 . title Pricing options with a new hybrid neural network model . journal Expert Systems with Applications volume 251 , pages 123979

    2024

  35. [43]

    , author Hong, L.J

    author Wang, S. , author Hong, L.J. , year 2021 . title Option pricing by neural stochastic differential equations: A simulation-optimization approach . journal 2021 Winter Simulation Conference (WSC) , pages 1--11

    2021

  36. [44]

    write newline

    " write newline " cite write " FUNCTION editor.postfix editor num.names #1 > "( )" "( )" if FUNCTION editor.trans.postfix editor num.names #1 > "( )" "( )" if FUNCTION trans.postfix translator num.names #1 > "( )" "( )" if FUNCTION authors.editors.reflist.apa5 'field := 'dot :...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.