Paper: arXiv 2402.00445

Authors: Takuji Arai, Yuto Imai

Abstract

This paper aims to develop a supervised deep-learning scheme to compute call option prices for the Barndorff-Nielsen and Shephard model with a non-martingale asset price process having infinite active jumps. In our deep learning scheme, teaching data is generated through the Monte Carlo method developed by Arai and Imai (2024). Moreover, the BNS model includes many variables, which makes the deep learning accuracy worse. Therefore, we will create another input variable using the Black-Scholes formula. As a result, the accuracy is improved dramatically.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 7.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper employs advanced stochastic calculus and Lévy processes for the BNS model, requiring deep mathematical formalism. However, it demonstrates strong empirical rigor by implementing Monte Carlo simulations for data generation, using Sobol sequences for quasi-random sampling, and reporting computational speed improvements (100x faster) in a deep learning framework.

Research Flowchart

  flowchart TD
  A["Research Goal: Compute BNS Call Option Prices via Supervised Deep Learning"] --> B["Generate Training Data<br>Monte Carlo Simulation Arai & Imai (2024)"]
  B --> C{"Input Variables"}
  C --> D["BNS Model Variables<br>High Dimensionality Issue"]
  C --> E["Black-Scholes Formula Input<br>Created to Improve Accuracy"]
  D & E --> F["Computational Process<br>Supervised Deep Learning Training"]
  F --> G["Key Findings<br>Dramatic Accuracy Improvement<br>Effective Option Pricing"]

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 supervised deep‑learning scheme to compute call option prices for the non‑martingale Barndorff‑Nielsen and Shephard model with infinite active jumps.

Source sentence
This paper aims to develop a supervised deep‑learning scheme to compute call option prices for the Barndorff‑Nielsen and Shephard model with a non‑martingale asset price process having infinite active jumps.

Assumes 5

  • Assumption 1.1 on model parameters ensures the minimal martingale measure is well‑defined. model
    Source sentence
    Assumption 1.1. Throughout this paper, we assume that b2/2 > 2 (1−e−λT)/λ ∨|ρ| and α e−λT σ2 0 + Cρ > −1.
  • The financial market consists of one risky asset and one riskless asset with zero interest rate. market
    Source sentence
    Throughout this paper, we consider a financial market with maturity T > 0, composed of one risky asset and one riskless asset with 0 interest rate.
  • The asset price process is non‑martingale under the physical measure. model
    Source sentence
    non‑martingale Barndorff‑Nielsen and Shephard (BNS) model with infinite active jumps.
  • Teaching data are generated by the Monte Carlo method of Arai and Imai (2024). data
    Source sentence
    In our deep learning scheme, teaching data is generated through the Monte Carlo method developed by Arai and Imai [2].
Show 1 more
  • The Black‑Scholes formula can be used as an additional input variable because BNS and Black‑Scholes option prices exhibit similar behavior. model
    Source sentence
    we can expect that option prices for the BNS model and the Black‑Scholes model exhibit similar behavior.

Data

  • Synthetic dataset of 100,000 samples synthetic unknown
    Source sentence
    In summary, we generate a dataset of 100,000 samples from an 8‑dimensional Sobol sequence. Each sample is composed of 10 variables: αn, ρn, λn, an, bn, (σ2 0)n, Kn, Tn, BSn and MCn.

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • The trained deep‑learning model computes option prices instantaneously, achieving roughly a 100‑fold speedup compared with Monte Carlo simulation (≈4 s vs. 178 h for 100 k samples). performance
    Source sentence
    perform option price computation using the Monte Carlo method alone, the trained deep learning model enables us to speed up by approximately 100 times. It takes about 4 seconds to implement one simulation with a regular laptop. However, the option price computation by the trained deep learning model is instantaneous.
  • Including the Black‑Scholes price as an input dramatically improves the accuracy of the deep‑learning scheme. predictive-accuracy
    Source sentence
    Therefore, we will create another input variable using the Black‑Scholes formula. As a result, the accuracy is improved dramatically.
  • Monte Carlo generation of 100,000 teaching samples is very time‑consuming (≈178 hours), limiting dataset size. other
    Source sentence
    The Monte Carlo method for 100,000 samples is very time‑consuming. It took 178 hours approximately.

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.

Code

No repository by the authors found in the paper.

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