Paper: arXiv 2412.02135

Authors: Changqing Teng, Guanglian Li

Abstract

Current deep learning-based calibration schemes for rough volatility models are based on the supervised learning framework, which can be costly due to a large amount of training data being generated. In this work, we propose a novel unsupervised learning-based scheme for the rough Bergomi (rBergomi) model which does not require accessing training data. The main idea is to use the backward stochastic differential equation (BSDE) derived in [“Bayer, Qiu and Yao, {“SIAM J. Financial Math.”}, 2022”] and simultaneously learn the BSDE solutions with the model parameters. We establish that the mean squares error between the option prices under the learned model parameters and the historical data is bounded by the loss function. Moreover, the loss can be made arbitrarily small under suitable conditions on the fitting ability of the rBergomi model to the market and the universal approximation capability of neural networks. Numerical experiments for both simulated and historical data confirm the efficiency of scheme.

Complexity vs Empirical Score

  • Math Complexity: 8.0/10
  • Empirical Rigor: 4.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper employs advanced stochastic calculus, backward stochastic differential equations (BSDEs), and rigorous convergence analysis, indicating high mathematical complexity. However, it focuses on a theoretical proof of concept with simulated and historical data validation, lacking the backtest-ready implementation details or datasets typical of high empirical rigor.

Research Flowchart

  flowchart TD
  A["Research Goal: Unsupervised calibration for Rough Bergomi model"] --> B["Input: Historical market option prices"]
  B --> C["Key Methodology: Formulate BSDE from Bayer, Qiu, Yao 2022"]
  C --> D["Compute: Loss function L = MSE of model vs. market prices"]
  D --> E["Optimize: Simultaneously learn Neural Network parameters and model parameters"]
  E --> F{"Check Convergence?"}
  F -- No --> D
  F -- Yes --> G["Output: Calibrated model parameters"]
  G --> H["Key Findings: <br>• Efficient scheme for synthetic/historical data<br>• Loss bounded by fitting error<br>• Loss approaches 0 under NN approximation"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper proposes a novel unsupervised learning-based calibration scheme for rough volatility models that eliminates the need for synthetic training data.

Source sentence
In this work, we propose a novel unsupervised learning-based calibration scheme for rough volatility models that eliminates the data generation bottleneck.

Assumes 3

  • The variance process V possesses a.s. continuous, non‑negative trajectories, is adapted to the filtration generated by the Brownian motion W, and is integrable. model
    Source sentence
    Assumption 1.1. The variance process V possesses a.s. continuous, non‑negative trajecto‑ries and is adapted to FW t . Furthermore, Vt is integrable; that is, E
  • The underlying price S0 is taken as the official closing price of the S&P 500 index and the spot variance V0 is proxied by the square of the at‑the‑money implied volatility with the shortest maturity. market
    Source sentence
    We designate the underlying price S0 as the official closing price of the S&P 500 index and proxy the spot variance V0 using the square of the at-the-money implied volatility with the shortest maturity.
  • Option prices are taken as the average of the maximum bid and minimum ask prices. data
    Source sentence
    The option prices used are the average of the maximum bid and minimum ask prices.

Data

  • S&P 500 European call option data · OptionMetrics via WRDS options commercial
    S&P 500 index (ticker: SPX)
    Source sentence
    We calibrate the model to S&P 500 index (ticker: SPX) European call option data sourced from OptionMetrics www.optionmetrics.com via Wharton Research Data Services (WRDS).
  • Synthetic data synthetic unknown
    Source sentence
    Numerical experiments for both simulated and historical S&P 500 data based on rough Bergomi (rBergomi) model demonstrate the efficiency and accuracy of the proposed scheme.

Recognised sources: OptionMetrics IvyDB WRDS

Costs

Not applicable The paper reports no trading or execution result.

Finds 4

  • The unsupervised calibration scheme achieves low calibration errors, e.g., average relative error as low as 1.2375e‑5 with step size 1/10. pricing-calibration
    Source sentence
    1/10 12.8426 10.5 1.2375e-5 2.4428e-6 0.0370 0.0049 0.0722 0.0130 0.0072
  • Historical calibration on SPX data yields calibrated underlying price S0 ≈ 3970.15 (Feb 28) and spot variance V0 ≈ 0.0378, with out‑of‑sample testing on unseen strikes and dates. pricing-calibration
    Source sentence
    Table 4: Underlying price S0 and spot variance V0 used for calibration.
  • Parameterizing the Hurst index as a time‑varying function improves both in‑sample and out‑of‑sample performance compared to scalar parameterizations. performance
    Source sentence
    parameterizing H(t) incurs higher computational costs but outperforms all other models on both in‑sample and out‑of‑sample datasets.
  • The proposed method matches or exceeds the accuracy of pure Monte‑Carlo calibration while avoiding the need for massive synthetic training datasets. theoretical
    Source sentence
    ... demonstrate the efficiency and accuracy of the proposed scheme.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 18 related papers). Coverage is partial, so this is neither replication nor refutation.

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.

Cited References (40)

  1. Convergence of the deep BSDE method for coupled FBSDEs (2020)
  2. TensorFlow : a system for large-scale machine learning (2016)
  3. Deep BSDE Solver in TensorFlow (2018)
  4. TF Quant Finance: TensorFlow-based Quantitative Finance Library (2020)
  5. Adam: A method for stochastic optimization (2015)
  6. On deep calibration of (rough) stochastic volatility models (2019) — link
  7. Deep hedging under rough volatility (2021)
  8. Deep learning volatility: a deep neural network perspective on pricing and calibration in rough volatility models (2021)
and 32 more
  1. Calibrating rough volatility models: a convolutional neural network approach (2020)
  2. Deep calibration of the quadratic rough Heston model (2021) — link
  3. A neural network-based framework for financial model calibration (2019)
  4. A generative adversarial network approach to calibration of local stochastic volatility models (2020)
  5. Model calibration with neural networks (2016)
  6. Backward stochastic differential equations in finance (1997)
  7. Deep learning-based numerical methods for high-dimensional parabolic partial differential equations and backward stochastic differential equations (2017)
  8. Solving high-dimensional partial differential equations using deep learning (2018)
  9. Efficient Calibration in the rough B ergomi model by W asserstein distance — → 2512.00448
  10. Pricing under rough volatility (2016)
  11. $ L ^p$ solutions of backward stochastic differential equations (2003)
  12. Volatility is rough (2018)
  13. Rough volatility, path-dependent PDEs and weak rates of convergence — → 2304.03042
  14. Hybrid scheme for Brownian semistationary processes (2017)
  15. A neural network warm-start approach for the inverse acoustic obstacle scattering problem (2023)
  16. Approximation rates for deep calibration of (rough) stochastic volatility models — → 2309.14784
  17. Deep backward schemes for high-dimensional nonlinear PDEs (2020)
  18. Discrete versions of G ronwall’s lemma and their application to the numerical analysis of parabolic problems (1999)
  19. Backward stochastic differential equations (2017)
  20. Is volatility rough? (2019) — link
  21. Asymptotics for rough stochastic volatility models (2017)
  22. Multifactor approximation of rough volatility models (2019)
  23. Volatility has to be rough (2021)
  24. A regularity structure for rough volatility (2020)
  25. Multifractional stochastic volatility models (2014)
  26. Rough volatility: evidence from option prices (2018)
  27. Empirical analysis of rough and classical stochastic volatility models to the SPX and VIX markets (2022)
  28. Short-time near-the-money skew in rough fractional volatility models (2019)
  29. On the martingale property in the rough B ergomi model (2019)
  30. Affine V olterra processes (2019)
  31. The characteristic function of rough H eston models (2019)
  32. Pricing options under rough volatility with backward SPDE s (2022)

3 of 40 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.