Paper: arXiv 2610.19254

Authors: Jose Miguel Lara Rangel, Mixtli Vazquez Mendoza

Abstract

Physics-informed neural networks provide a flexible framework for learning parametric PDE solution families without labeled data, yet preserving accurate derivatives across heterogeneous regimes remains challenging. Structured neural solvers offer compact representations, but restrictive conditional parameterizations can limit adaptation across the solution family. We introduce BeTINN, a structured conditional architecture extending time-induced modulation through multi-directional time-parameter conditioning and a joint residual pathway. We evaluate BeTINN on the Heston stochastic-volatility PDE, a demanding parametric problem with mixed derivatives and economically relevant first- and second-order sensitivities. BeTINN improves joint accuracy across normalized option value and key sensitivity metrics relative to our baselines. Controlled ablations characterize the roles of its architectural components. We further find that validation physics loss is informative but does not fully determine differential fidelity, and that accurate solution values can coexist with localized derivative errors. These results support structured conditional adaptation and motivate derivative-aware evaluation when sensitivities are downstream quantities of interest.

Complexity vs Empirical Score

  • Math Complexity: 7.5/10
  • Empirical Rigor: 6.5/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: This paper presents a novel neural network architecture for solving parametric PDEs in finance, specifically the Heston model. It demonstrates strong mathematical foundations and a solid empirical evaluation of its performance against baselines, focusing on both solution accuracy and derivative fidelity. The work is innovative in its approach to conditional adaptation for neural solvers.

Research Flowchart

  flowchart TD
    A[Research Goal: Improve Parametric PDE Solving with PINNs] --> B{Methodology: BeTINN Architecture};
    B --> C[Inputs: Heston Stochastic-Volatility PDE & Parameters];
    C --> D{Computational Process: Training & Evaluation of BeTINN};
    D --> E[Outcome 1: Improved Joint Accuracy in Option Value & Sensitivities];
    D --> F[Outcome 2: Validation Physics Loss ≠ Full Derivative Fidelity];
    D --> G[Outcome 3: Accurate Values can Coexist with Localized Derivative Errors];

Cited References (24)

  1. DGM: A deep learning algorithm for solving partial differential equations (2018)
  2. The Heston model and its extensions in Matlab and C# (2013)
  3. Physics informed deep learning (part i): Data-driven solutions of nonlinear partial differential equations (2017) — link
  4. TINNs: Time-Induced Neural Networks for Solving Time-Dependent PDEs (2026) — link
  5. DeepSVM: Learning Stochastic Volatility Models with Physics-Informed Deep Operator Networks — → 2512.07162
  6. Option pricing in the Heston model with physics inspired neural networks (2024)
  7. A metalearning approach for physics-informed neural networks (PINNs): Application to parameterized PDEs (2023)
  8. HyperPINN: Learning parameterized differential equations with physics-informed hypernetworks (2021)
and 16 more
  1. Parameterized physics-informed neural networks for parameterized PDEs (2024) — link
  2. Conditionally parameterized, discretization-aware neural networks for mesh-based modeling of physical systems (2021)
  3. Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators (2021)
  4. Fourier neural operator for parametric partial differential equations (2020) — link
  5. Learning the solution operator of parametric partial differential equations with physics-informed DeepONets (2021)
  6. Sobolev training for neural networks (2017)
  7. Gradient-enhanced physics-informed neural networks for forward and inverse PDE problems (2022)
  8. A closed-form solution for options with stochastic volatility with applications to bond and currency options (1993)
  9. Physics-informed machine learning (2021)
  10. Neural operator: Learning maps between function spaces with applications to pdes (2023)
  11. Automatic differentiation in machine learning: a survey (2018)
  12. Differential machine learning (2020) — link
  13. Calibrating the Heston model with deep differential networks — → 2407.15536
  14. Neural Operator Processes for Probabilistic Operator Learning under Partial Observations (2026) — link
  15. The Levenberg-Marquardt algorithm: implementation and theory (2006)
  16. Hypernetwork-based meta-learning for low-rank physics-informed neural networks (2023)

2 of 24 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.