Paper: arXiv 2407.13688

Authors: Nacira Agram, Bernt Øksendal, Jan Rems

Abstract

We propose a deep learning approach to study the minimal variance pricing and hedging problem in an incomplete jump diffusion market. It is based upon a rigorous stochastic calculus derivation of the optimal hedging portfolio, optimal option price, and the corresponding equivalent martingale measure through the means of the Stackelberg game approach. A deep learning algorithm based on the combination of the feedforward and LSTM neural networks is tested on three different market models, two of which are incomplete. In contrast, the complete market Black-Scholes model serves as a benchmark for the algorithm’s performance. The results that indicate the algorithm’s good performance are presented and discussed. In particular, we apply our results to the special incomplete market model studied by Merton and give a detailed comparison between our results based on the minimal variance principle and the results obtained by Merton based on a different pricing principle. Using deep learning, we find that the minimal variance principle leads to typically higher option prices than those deduced from the Merton principle. On the other hand, the minimal variance principle leads to lower losses than the Merton principle.

Complexity vs Empirical Score

  • Math Complexity: 8.0/10
  • Empirical Rigor: 5.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: The paper is dense with advanced stochastic calculus, PDE/BSDE formulations, and rigorous theoretical derivations for hedging in jump-diffusion models, indicating high mathematical complexity. It employs deep learning (LSTM/feedforward networks) and tests on multiple market models (Black-Scholes, multi-dimensional Brownian, Merton) with explicit performance discussions, demonstrating significant empirical implementation.

Research Flowchart

  flowchart TD
  Start["Research Goal: <br>Deep Learning for Minimal<br>Variance Hedging in<br>Incomplete Jump Markets"] --> Methodology["Key Methodology:<br>Stackelberg Game Approach<br>+<br>Feedforward & LSTM Networks"]
  Methodology --> Inputs["Data/Inputs:<br>1. Black-Scholes (Benchmark)<br>2. Incomplete Jump Diffusion<br>3. Merton Model"]
  Inputs --> Process["Computational Process:<br>Training Neural Networks<br>to Optimize Portfolio<br>and Pricing Function"]
  Process --> Results["Key Findings/Outcomes:<br>1. Algorithm performs well<br>in incomplete markets<br>2. Minimal Variance prices<br>higher than Merton principle<br>3. Lower hedging losses<br>than Merton principle"]

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 learning approach to solve the minimal variance pricing and hedging problem in an incomplete jump diffusion market.

Source sentence
We propose a deep learning approach to study the minimal variance pricing and hedging problem in an incomplete jump diffusion market.

Assumes 6

  • The market is modeled as a jump diffusion market. market
    Source sentence
    We propose a deep learning approach to study the minimal variance pricing and hedging problem in an incomplete jump diffusion market.
  • The volatility σ is assumed constant and the jump size function γ(t,ζ) does not depend on time. model
    Source sentence
    assuming for simplicity that σ is constant and γ(t, ζ) = γ(ζ) does not depend on t
  • The underlying asset may depend on multiple independent Brownian motions in the continuous incomplete market case. model
    Source sentence
    a continuous model where the underlying asset depends on multiple independent Brownian motions is used
  • No calibration of SDEs to market data is performed. data
    Source sentence
    Furthermore, we do not consider the calibration of SDEs to fit market data in this paper.
Show 2 more
  • Synthetic data are generated by discretizing the time interval and sampling Brownian and Poisson increments. data
    Source sentence
    To generate the data, we discretize time interval [0, T] into R equidistant points ti at a distance ∆t.
  • The Black‑Scholes model is used as a complete‑market benchmark where option price and hedging portfolio are known analytically. other
    Source sentence
    test this algorithm in the case of a complete market, namely in the case of the Black Scholes (BS) model.

Data

  • Synthetic simulated paths synthetic public
    Source sentence
    To generate the data, we discretize time interval [0, T] into R equidistant points ti at a distance ∆t.

Costs

Unclear No cost treatment was found in the passages the extractor read; that is not evidence that costs are not modeled.

Finds 4

  • The deep learning algorithm shows good performance on all three tested models, including the incomplete Merton and Kou jump‑diffusion markets. performance
    Source sentence
    We show the algorithm’s success in all three models and discuss its performance.
  • Using the minimal variance principle yields typically higher option prices than those obtained from Merton’s pricing principle. pricing-calibration
    Source sentence
    Using deep learning, we find that the minimal variance principle leads to typically higher option prices than those deduced from the Merton principle.
  • The minimal variance principle leads to lower hedging losses compared with the Merton principle. performance
    Source sentence
    On the other hand, the minimal variance principle leads to lower losses than the Merton principle.
  • Option price comparisons show that price differences depend on parameters λ, µ, σ and δ, with the biggest differences when jump sizes are constant and volatility is large. performance
    Source sentence
    prices seems to be biggest when jump sizes are constant while volatility in the model is large.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 30 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

https://github.com/janrems/DeepLearningQHedging — the authors' own repository

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