Paper: arXiv 2502.11706
Authors: Balint Negyesi, Cornelis W. Oosterlee
Abstract
A deep BSDE approach is presented for the pricing and delta-gamma hedging of high-dimensional Bermudan options, with applications in portfolio risk management. Large portfolios of a mixture of multi-asset European and Bermudan derivatives are cast into the framework of discretely reflected BSDEs. This system is discretized by the One Step Malliavin scheme (Negyesi et al. [“2024, 2025”]) of discretely reflected Markovian BSDEs, which involves a \(Γ\) process, corresponding to second-order sensitivities of the associated option prices. The discretized system is solved by a neural network regression Monte Carlo method, efficiently for a large number of underlyings. The resulting option Deltas and Gammas are used to discretely rebalance the corresponding replicating strategies. Numerical experiments are presented on both high-dimensional basket options and large portfolios consisting of multiple options with varying early exercise rights, moneyness and volatility. These examples demonstrate the robustness and accuracy of the method up to \(100\) risk factors. The resulting hedging strategies significantly outperform benchmark methods both in the case of standard delta- and delta-gamma hedging.
Complexity vs Empirical Score
- Math Complexity: 9.2/10
- Empirical Rigor: 8.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced stochastic calculus including BSDEs, Malliavin calculus, and high-dimensional neural network approximations, representing a very high math complexity. The method is demonstrated with numerical experiments on portfolios up to 100 risk factors, showing backtest-ready implementation with empirical validation of hedging performance.
Research Flowchart
flowchart TD A["Research Goal"] --> B["Problem Formulation"] B --> C["Discretization"] C --> D["Computational Engine"] D --> E["Outcomes & Validation"] A["Research Goal<br>Simultaneous Pricing & Delta-Gamma Hedging<br>High-Dim Bermudan Portfolios"] B["Problem Formulation<br>Large portfolios of multi-asset options<br>Cast as Discretely Reflected BSDEs"] C["Discretization<br>One Step Malliavin Scheme<br>(Includes Γ-process for Gamma)"] D["Computational Engine<br>Neural Network Regression<br>Monte Carlo Simulation"] E["Outcomes & Validation<br>Accurate Deltas/Gammas up to 100 dimensions<br>Superior hedging performance<br>vs. benchmarks"]
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 deep BSDE methodology for simultaneous pricing, delta and delta‑gamma hedging of large high‑dimensional Bermudan option portfolios.
Source sentence
We propose a deep BSDE methodology for the portfolio hedging problem outlined above.
Assumes 3
- Options can only be exercised on rebalancing dates. execution
Source sentence
In each example below, we assume that options can only be exercised on rebalancing dates.
- The time step size Δtn is small or the volatility is high for the scheme to work well. model
Source sentence
the time step size ∆tn is small or the volatility is high.
- Underlying deltas are highly volatile, which makes delta‑gamma hedging advantageous. market
Source sentence
whenever the underlyings’ deltas are highly volatile, resulting in less frequent rebalancing, and sharper Profit-and-Loss (PnL) distributions.
Data
- High‑dimensional basket options and large synthetic portfolios synthetic unknown
Source sentence
Numerical experiments are presented on both high-dimensional basket options and large portfolios consisting of multiple options with varying early exercise rights, moneyness and volatility.
Costs
Not modeled
Finds 3
- Gamma hedging significantly outperforms delta hedging in replication accuracy, achieving the same accuracy with far fewer rebalancing dates. performance
Source sentence
we see that Gamma hedging significantly outperforms delta hedging in terms of replication accuracy, irrespective of the moneyness.
- The One Step Malliavin (OSM) scheme yields more accurate Delta approximations, especially under high volatility and small time steps. performance
Source sentence
the OSM scheme results in gradually more accurate Delta approximations compared to the reference method [18], which results in sharper PnL‑distributions.
- Delta‑gamma hedging reaches a risk‑tolerance level of VaR95 ≤10% for rebalancing frequencies higher than N = 10, whereas pure delta hedging does not achieve this even with daily rebalancing. performance
Source sentence
the delta‑gamma hedging strategy achieves a risk tolerance level of VaR95 ≤10%.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 53 related papers). Coverage is partial, so this is neither replication nor refutation.
4 papers in the archive : KANHedge: Efficient Hedging of High-Dimensional Options …
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (50)
- A comparison of biased simulation schemes for stochastic volatility models (2010) — link
- Two algorithms for the discrete time approximation of Markovian backward stochastic differential equations under local conditions (2015) — link
- Convergence of Monte Carlo simulations involving the mean-reverting square root process (2005) — link
- Representations and regularities for solutions to BSDEs with reflections (2005) — link
- Discrete-time approximation for continuously and discretely reflected BSDEs (2008) — link
- Reflected solutions of backward SDE ’s, and related obstacle problems for PDE ’s (1997) — link
- DGM : A deep learning algorithm for solving partial differential equations (2018) — link
- The Tracking Error Rate of the Delta - Gamma Hedging Strategy (2012) — link
and 42 more
- An Introduction to Computational Finance Without Agonizing Pain
- Backward Stochastic Differential Equations (2017) — link
- Deep Runge - Kutta schemes for BSDEs — → 2212.14372
- The One Step Malliavin scheme: new discretization of BSDEs implemented with deep learning regressions (2021) — link
- Deep learning for CVA computations of large portfolios of financial derivatives (2021) — link
- Brownian Motion and Stochastic Calculus (1998) — link
- The Malliavin Calculus and Related Topics (2006) — link
- Deep backward multistep schemes for nonlinear PDEs and approximation error analysis (2020)
- Path regularity and explicit convergence rate for BSDE with truncated quadratic growth (2010) — link
- Numerical simulation of quadratic BSDEs (2016) — link
- Efficient numerical Fourier methods for coupled forward–backward SDEs (2016) — link
- Deep neural network framework based on backward stochastic differential equations for pricing and hedging American options in high dimensions (2021) — link
- Deep backward schemes for high-dimensional nonlinear PDEs (2020) — link
- Backward stochastic differential equations and quasilinear parabolic partial differential equations (1992) — link
- A regression-based Monte Carlo method to solve backward stochastic differential equations (2005) — link
- A quantization algorithm for solving multidimensional discrete-time optimal stopping problems (2003) — link
- A Fourier Cosine Method for an Efficient Computation of Solutions to BSDEs (2015) — link
- A forward scheme for backward SDEs (2007) — link
- Computation of Greeks for Barrier and Lookback Options Using Malliavin Calculus (2003) — link
- Hedging of the European Option of the Exotic Type with a Nonsmooth Payoff Function (2017)
- Applications of Malliavin calculus to Monte Carlo methods in finance (1999) — link
- Discrete-time approximation and Monte - Carlo simulation of backward stochastic differential equations (2004) — link
- Pricing and Hedging American - Style Options with Deep Learning (2020) — link
- Two- Dimensional Fourier Cosine Series Expansion Method for Pricing Financial Options (2012) — link
- Universal approximation of an unknown mapping and its derivatives using multilayer feedforward networks (1990) — link
- Solving high-dimensional partial differential equations using deep learning (2018) — link
- Solving forward-backward stochastic differential equations explicitly — a four step scheme (1994) — link
- Simulation of BSDEs by Wiener chaos expansion (2014) — link
- Reflected Backward SDEs (2017) — link
- Convergence of the deep BSDE method for coupled FBSDEs (2020) — link
- Backward Stochastic Differential Equations in Finance (1997) — link
- Malliavin calculus for backward stochastic differential equations and application to numerical solutions (2011) — link
- Pricing early-exercise and discrete barrier options by fourier-cosine series expansions (2009) — link
- Reflected extension to the One Step Malliavin scheme and its corresponding deep BSDE approximations (working paper) (2024)
- The One Step Malliavin scheme: new discretization of BSDEs implemented with deep learning regressions (2024) — link
- The Value of an Option to Exchange One Asset for Another (1978) — link
- LSQR : An Algorithm for Sparse Linear Equations and Sparse Least Squares (1982) — link
- Algorithm 583: LSQR : Sparse Linear Equations and Least Squares Problems (1982) — link
- Continuous Time Wishart Process for Stochastic Risk (2006) — link
- Efficient Simulation of the Wishart Model (2009) — link
- Approximation Error Analysis of Some Deep Backward Schemes for Nonlinear PDEs (2022) — link
- Deep Quadratic Hedging — → 2212.12725
2 of 50 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.