Paper: arXiv 2309.10477
Authors: Pierre-Antoine Arsaguet, Paul Bilokon
Abstract
This report investigates the computation of option Greeks for European and Asian options under the Heston stochastic volatility model on GPU. We first implemented the exact simulation method proposed by Broadie and Kaya and used it as a baseline for precision and speed. We then proposed a novel method for computing Greeks using the Milstein discretisation method on GPU. Our results show that the proposed method provides a speed-up up to 200x compared to the exact simulation implementation and that it can be used for both European and Asian options. However, the accuracy of the GPU method for estimating Rho is inferior to the CPU method. Overall, our study demonstrates the potential of GPU for computing derivatives sensitivies with numerical methods.
Complexity vs Empirical Score
- Math Complexity: 8.5/10
- Empirical Rigor: 6.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper is heavy on advanced mathematics, featuring detailed SDE formulations, Monte Carlo simulation algorithms (Broadie and Kaya), and specific discretization methods (Milstein), requiring significant theoretical understanding. Empirically, it involves concrete GPU implementation, performance benchmarking (200x speed-up), and accuracy comparisons against CPU baselines for derivatives sensitivities, making it directly applicable to quantitative finance engineering.
Research Flowchart
flowchart TD
A["Research Goal: Compute Heston Model Greeks for European & Asian Options on GPU"] --> B["Data & Inputs"]
B --> C["Methodology: Baseline CPU & Proposed GPU"]
subgraph C ["Methodology"]
C1["Baseline: Broadie-Kaya<br>Exact Simulation CPU"]
C2["Proposed: Milstein Discretisation<br>on GPU"]
end
C --> D["Computational Process: Parallel Simulation & Greeks Calculation"]
D --> E["Key Findings & Outcomes"]
subgraph E ["Outcomes"]
E1["200x Speed-up vs Baseline"]
E2["Valid for European & Asian Options"]
E3["Limitation: Rho Accuracy < CPU"]
end
%% Data/Inputs Node Details
B1["Market Data: Spot, Strike, T, r, q"] --> B
B2["Heston Parameters: kappa, theta, sigma, rho"] --> B
B3["Simulation Parameters: Maturity, Paths, Steps"] --> B
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 method for computing option Greeks using the Milstein discretisation on GPU, called MGPU, achieving large speedups over exact simulation.
Source sentence
We then proposed a novel method for computing Greeks using the Milstein discretisation method on GPU.
Assumes 6
- The underlying asset price follows the Heston stochastic volatility model. model
Source sentence
Then, we need a model to estimate the asset price at time T. This could be done by models such as Black‑Scholes model, or Heston model, which is used for this work.
- Pricing is performed under the risk‑neutral measure with a constant risk‑free rate. market
Source sentence
The price of the option at time t is given by the risk‑neutral expectation of the discounted payoff, that is, V (t, St) = e^{‑r(T‑t)}E^Q [h(ST ) | St] , where r is the risk‑free interest rate.
- Monte Carlo simulation is used to generate asset paths. execution
Source sentence
Assume that we want to use Monte Carlo method to price an European option under Heston model.
- Correlated normal random variables are generated using a standard linear transformation. statistical
Source sentence
Z1, Z2 are correlated by ρ. To generate two correlated Normal random variable, one can use the algorithm 2.4.1.
Show 2 more
- GPU parallelism can be exploited for path‑wise simulations. execution
Source sentence
GPUs are particularly good at parallelising such simulations and thus accelerating them.
- Euler discretisation is less precise but computationally efficient. model
Source sentence
Finally, although Euler discretisation is less precise, it has the advantage of being computationally efficient and easily parallelisable.
Data
- Heston model parameters (Table 1) synthetic unknown
Source sentence
Variable κ θ σ v0 T r S0 K Value 6.21 0.019 0.61 0.010201 1 0.0319 100 100
Costs
Not applicable
Finds 5
- The MGPU implementation achieves speed‑ups of up to 200× compared with the exact simulation CPU baseline. performance
Source sentence
Our results show that the proposed method provides a speed‑up up to 200x compared to the exact simulation implementation
- GPU estimation of Rho is less accurate than the CPU exact simulation, with MGPU Rho deviating from the exact value. negative
Source sentence
the accuracy of the GPU method for estimating Rho is inferior to the CPU method
- MGPU delivers significant speedup especially for Asian option price computation over the exact simulation CPU implementation. performance
Source sentence
Our implementation, MGPU, achieved significant speedup over the Exact Simulation CPU implementation, especially for the computation of Asian option prices.
- Quasi‑Monte Carlo methods did not improve accuracy in the high‑dimensional setting for Milstein and Euler discretisations. negative
Source sentence
the results were unsatisfactory in the high‑dimensional setting
- Random number generation within each CUDA kernel adds to the total execution time of a path in the MGPU implementation. other
Source sentence
Notably, we generated random variables within each kernel, so the total execution time of a path under the MGPU implementation includes random variable generation.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 49 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.