Paper: arXiv 2402.06840
Authors: Duy-Minh Dang, Hao Zhou
Abstract
Option contracts on two underlying assets within uncertain volatility models have their worst-case and best-case prices determined by a two-dimensional (2D) Hamilton-Jacobi-Bellman (HJB) partial differential equation (PDE) with cross-derivative terms. This paper introduces a novel ``decompose and integrate, then optimize’’ approach to tackle this HJB PDE. Within each timestep, our method applies piecewise constant control, yielding a set of independent linear 2D PDEs, each corresponding to a discretized control value. Leveraging closed-form Green’s functions, these PDEs are efficiently solved via 2D convolution integrals using a monotone numerical integration method. The value function and optimal control are then obtained by synthesizing the solutions of the individual PDEs. For enhanced efficiency, we implement the integration via Fast Fourier Transforms, exploiting the Toeplitz matrix structure. The proposed method is \(\ell_{"\infty"}\)-stable, consistent in the viscosity sense, and converges to the viscosity solution of the HJB equation. Numerical results show excellent agreement with benchmark solutions obtained by finite differences, tree methods, and Monte Carlo simulation, highlighting its robustness and effectiveness.
Complexity vs Empirical Score
- Math Complexity: 9.0/10
- Empirical Rigor: 7.0/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper presents advanced mathematics involving high-dimensional HJB PDEs, viscosity solutions, and analytical Green’s functions, warranting a high math complexity score. While the empirical rigor is bolstered by extensive numerical comparisons with benchmark methods and robust convergence proofs, the absence of raw datasets or explicit backtesting procedures keeps it below a perfect score.
Research Flowchart
flowchart TD A["Research Goal: Solve 2D HJB PDE<br>for uncertain volatility models"] --> B["Key Methodology:<br>'Decompose & Integrate, then Optimize'"] B --> C["Data/Inputs:<br>Option Specs &<br>Uncertain Volatility Bounds"] C --> D["Computational Process:<br>1. Piecewise Constant Control<br>2. 2D Convolution (Green's Functions)<br>via Fast Fourier Transform (FFT)"] D --> E["Key Findings/Outcomes:<br>- ℓ∞-stable & Viscosity Consistent<br>- Fast FFT-based implementation<br>- Accurate vs Benchmarks<br>(FD, Trees, Monte Carlo)"]
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 “decompose and integrate, then optimize” method using piecewise constant control and Green's function convolution to solve the 2D HJB PDE for two‑factor uncertain volatility models.
Source sentence
This paper introduces a novel ``decompose and integrate, then optimize'' approach to tackle this HJB PDE.
Assumes 5
- The correlation coefficient between the two underlying assets lies strictly between –1 and 1. model
Source sentence
our focus remains predominantly on scenarios where the correlation coefficient lies strictly between −1 and 1
- The absolute correlation satisfies |ρ| < 1 for the theoretical analysis of the Green’s function truncation error. model
Source sentence
Suppose that |ρ| < 1, and let ∆τ > 0 be fixed.
- Piecewise constant control is applied within each time step, yielding independent linear 2D PDEs for each discretized control value. model
Source sentence
which yields a set of independent 2D linear PDEs, each corresponding to a fixed control value.
- The Dirac delta function in the formulation is approximated by a Gaussian function for computational purposes. model
Source sentence
For computational purposes, approximating the Dirac delta function δ(y −(a ± bx)) in (3.13) by a suitable Gaussian function is necessary
Show 1 more
- The computational spatial domain is chosen based on an upper bound for the boundary truncation error of the Green’s function. other
Source sentence
This approach balances the need for demonstrating theoretical convergence and computational efficiency in our analysis.
Data
Synthetic data only.
Costs
Not applicable
Finds 5
- Numerical results show remarkable agreement with benchmark solutions from monotone finite differences, tree‑grid methods, and Monte Carlo simulation. theoretical
Source sentence
Extensive numerical results show remarkable agreement with benchmark solutions from monotone FD, tree‑grid methods, and Monte Carlo simulation
- For convex payoffs the method attains high accuracy with a single time step, outperforming unconditionally monotone FD schemes in both accuracy and run time. performance
Source sentence
for convex payoffs, our method achieves high accuracy with a single time step, and even in general settings with multiple time steps, it significantly outperforms unconditionally monotone FD schemes in both accuracy and run time
- The scheme exhibits experimentally first‑order convergence for general payoffs, exceeding the previously proved 1/6 rate. statistical
Source sentence
we often observe experimentally first-order convergence, significantly exceeding the 1/6 rate proved in [28]
- The method is ℓ∞‑stable, consistent in the viscosity sense, and converges pointwise to the viscosity solution of the HJB equation. theoretical
Source sentence
We have mathematically demonstrated the ℓ∞‑stability and consistency of the scheme in the viscosity sense, along with its pointwise convergence to the viscosity solution of the HJB equation
- Implementation using FFTs and circulant convolution exploits Toeplitz structure to accelerate both inner and double summations, yielding significant run‑time improvements. performance
Source sentence
we have also implemented our scheme efficiently using FFT and circulant convolution, exploiting the Toeplitz structure of the convolution kernels to accelerate both inner and double summations via 2D FFTs
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 32 related papers). Coverage is partial, so this is neither replication nor refutation.
1 paper in the archive
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.