Paper: arXiv 2506.14612

Authors: Howard Su, Huan-Hsin Tseng

Abstract

We propose a quantum machine learning framework for approximating solutions to high-dimensional parabolic partial differential equations (PDEs) that can be reformulated as backward stochastic differential equations (BSDEs). In contrast to popular quantum-classical network hybrid approaches, this study employs the pure Variational Quantum Circuit (VQC) as the core solver without trainable classical neural networks. The quantum BSDE solver performs pathwise approximation via temporal discretization and Monte Carlo simulation, framed as model-based reinforcement learning. We benchmark VQCbased and classical deep neural network (DNN) solvers on two canonical PDEs as representatives: the Black-Scholes and nonlinear Hamilton-Jacobi-Bellman (HJB) equations. The VQC achieves lower variance and improved accuracy in most cases, particularly in highly nonlinear regimes and for out-of-themoney options, demonstrating greater robustness than DNNs. These results, obtained via quantum circuit simulation, highlight the potential of VQCs as scalable and stable solvers for highdimensional stochastic control problems.

Complexity vs Empirical Score

  • Math Complexity: 9.0/10
  • Empirical Rigor: 3.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper is densely packed with advanced mathematics, including deep stochastic calculus, PDE theory, and quantum circuit formulations, with rigorous theoretical derivations. However, the empirical results are limited to numerical simulations without backtesting, real market data, or implementation details for live trading, placing it firmly in the theoretical research domain.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Develop a pure Variational Quantum Circuit<br>solver for high-dimensional PDEs"] --> B["Methodology: Quantum BSDE Solver"]
  
  B --> C["Data: PDE Benchmarks"]
  C --> C1["Black-Scholes Eq.<br>Linear"]
  C --> C2["Nonlinear HJB Eq.<br>High-dim control"]
  
  B --> D["Key Processes"]
  D --> D1["Pathwise Approx.<br>via Temporal Discretization"]
  D --> D2["Model-Based RL<br>with Monte Carlo Paths"]
  
  B --> E["Computational Core<br>Pure VQC (No Classical NNs)"]
  
  E --> F["Outcomes vs Classical DNN"]
  
  F --> G["Lower Variance<br>(Better Stability)"]
  F --> H["Improved Accuracy<br>(Nonlinear & Out-of-Money)"]
  
  subgraph Legend
      direction LR
      L1["Research Goal"]
      L2["Methodology"]
      L3["Data/Inputs"]
      L4["Process"]
      L5["Outcome"]
  end
  
  style A fill:#e1f5e1,stroke:#2e7d32,stroke-width:2px
  style B fill:#e3f2fd,stroke:#1565c0,stroke-width:2px
  style C fill:#fff3e0,stroke:#ef6c00,stroke-width:2px
  style C1 fill:#fff3e0,stroke:#ef6c00
  style C2 fill:#fff3e0,stroke:#ef6c00
  style D fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px
  style D1 fill:#f3e5f5,stroke:#7b1fa2
  style D2 fill:#f3e5f5,stroke:#7b1fa2
  style E fill:#e8f5e9,stroke:#1b5e20,stroke-width:2px
  style F fill:#fce4ec,stroke:#c2185b,stroke-width:2px
  style G fill:#fce4ec,stroke:#c2185b
  style H fill:#fce4ec,stroke:#c2185b
  style L1 fill:#e1f5e1,stroke:#2e7d32
  style L2 fill:#e3f2fd,stroke:#1565c0
  style L3 fill:#fff3e0,stroke:#ef6c00
  style L4 fill:#f3e5f5,stroke:#7b1fa2
  style L5 fill:#fce4ec,stroke:#c2185b

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 pure variational quantum circuit (VQC) based BSDE solver as a quantum machine learning framework for approximating high‑dimensional parabolic PDE solutions.

Source sentence
We propose a quantum machine learning framework for approximating solutions to high-dimensional parabolic partial differential equations (PDEs) that can be reformulated as backward stochastic differential equations (BSDEs).

Assumes 2

  • The functions b, σ, f, and g are measurable and satisfy a linear growth bound. model
    Source sentence
    The functions b, σ, f, and g are measurable, and there exists C > 0 such that |b(x)|+∥σ(x)∥+|f(t, x, y, z)|+|g(x)| ≤C(1+|x|+|y|+|z|), for all (t, x, y, z) ∈[0, T] × Rd × R × Rd.
  • The driver f(t,x,y,z) is Lipschitz continuous in (y,z) uniformly in (t,x). model
    Source sentence
    The function f(t, x, y, z) is Lipschitz continuous in (y, z) uniformly in (t, x).

Data

  • Black–Scholes equation synthetic public
    Source sentence
    Both examples admit closed-form analytical solutions, enabling rigorous error assessment.
  • Nonlinear Hamilton‑Jacobi‑Bellman (HJB) equation synthetic public
    Source sentence
    We benchmark the method on two representative parabolic PDEs: the Black‑Scholes equation and a nonlinear Hamilton‑Jacobi‑Bellman (HJB) equation.

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • The VQC solver shows consistently lower variance than the DNN solver for call option pricing. performance
    Source sentence
    The VQC and DNN produce similar average results, but the variance of the VQC errors is consistently lower.
  • For out‑of‑the‑money put options, the VQC outperforms the DNN in both average accuracy and error variance. performance
    Source sentence
    In this case, the VQC outperforms the DNN in both average accuracy and error variance.
  • Across all runs on the HJB equation, the VQC‑based solver consistently achieves lower average error and smaller variance compared to the DNN. performance
    Source sentence
    Across all runs, the VQC-based solver consistently achieves lower average error and smaller variance compared to the DNN.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 28 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.

Cited References (14)

  1. An overview on deep learning-based approximation methods for partial differential equations (2022)
  2. Three ways to solve partial differential equations with neural networks–A review (2021)
  3. Deep BSDE method for solving high-dimensional partial differential equations (2017)
  4. Backward stochastic differential equations and quasilinear parabolic partial differential equations (1992)
  5. Forward-backward stochastic differential equations and quasilinear parabolic PDEs (1999)
  6. Solving high-dimensional partial differential equations using deep learning (2018)
  7. Variational quantum algorithms (2021)
  8. A variational eigenvalue solver on a photonic quantum processor (2014) — link
and 6 more
  1. The theory of variational hybrid quantum–classical algorithms (2016) — link
  2. A Quantum Approximate Optimization Algorithm (2014) — link
  3. Quantum Approximate Optimization Algorithm: Performance, Mechanism, and Implementation on Near-Term Devices (2020) — link
  4. Quantum circuit learning (2018)
  5. Variational quantum circuits for deep reinforcement learning (2020)
  6. Quantum long short-term memory (2022)

0 of 14 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.