Paper: arXiv 2410.10474

Authors: Naman Krishna Pande, Puneet Pasricha, Arun Kumar, Arvind Kumar Gupta

Abstract

In this article, we employ physics-informed residual learning (PIRL) and propose a pricing method for European options under a regime-switching framework, where closed-form solutions are not available. We demonstrate that the proposed approach serves an efficient alternative to competing pricing techniques for regime-switching models in the literature. Specifically, we demonstrate that PIRLs eliminate the need for retraining and become nearly instantaneous once trained, thus, offering an efficient and flexible tool for pricing options across a broad range of specifications and parameters.

Complexity vs Empirical Score

  • Math Complexity: 8.0/10
  • Empirical Rigor: 3.0/10
  • Quadrant: Lab Rats — high math complexity, limited empirical testing

Why this score: The paper introduces advanced mathematics through regime-switching models and physics-informed neural networks (PIRL), with heavy derivations and PDE formulations. However, it lacks empirical data, backtests, or code, focusing instead on theoretical methodology and numerical experiments within a controlled academic setting.

Research Flowchart

  flowchart TD
  A["Research Goal: European Option Pricing in Regime-Switching Models"] --> B["Key Methodology: Physics-Informed Residual Learning PIRL"]
  B --> C["Data Inputs: Market Data & Regime Probabilities"]
  B --> D["Computational Process: Train PIRL to satisfy Black-Scholes PDE"]
  C --> D
  D --> E["Outcome 1: No closed-form solution required"]
  D --> F["Outcome 2: Near-instantaneous pricing post-training"]
  D --> G["Outcome 3: Flexible across parameters & regimes"]

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 physics-informed residual learning (PIRL) pricing method for European options in a regime‑switching framework.

Source sentence
we employ physics-informed residual learning (PIRL) and propose a pricing method for European options under a regime‑switching framework, where closed‑form solutions are not available.

Assumes 5

  • The financial market is frictionless with a cash account earning a constant risk‑free rate and one underlying asset. market
    Source sentence
    (a) a cash account that accrues interest at a constant risk‑free rate r, and (b) an underlying asset with price at time t denoted by St.
  • The regime process X_t is a continuous‑time Markov chain with two states (bullish and bearish). model
    Source sentence
    We, further, assume that Xt is a continuous time Markov chain (CTMC) with two states
  • State transitions follow a Poisson process with intensity λ_jk. model
    Source sentence
    Since Xt is a CTMC, the transition between the two states occurs as a Poisson process, i.e., P(t*_{jk} > t) = e^{‑λ_{jk}t}, j, k = 1, 2, j̸= k.
  • The risk related to a state transition is diversifiable and therefore unpriced. model
    Source sentence
    Following Zhu et al. (2012), we make the assumption that the risk related to a state transition is diversifiable and therefore unpriced, an assumption that does not compromise generality.
Show 1 more
  • The underlying Wiener process W_t and the regime process X_t are independent. model
    Source sentence
    the process {W_t} is the standard Wiener process and we assume that the processes X and W are independent.

Data

  • Training data for BSM regime‑switching model synthetic unknown
    Source sentence
    Parameter Range… S0 [40, 100] r [0.01, 0.025] σ1 [0.10, 0.30] σ2 [σ1, 0.40] T [0, 4] t [0, T] Table 1: Sampling ranges for the parameters for generating the training data for BSM regime‑switching model.

Costs

Not applicable The paper reports no trading or execution result.

Finds 4

  • PIRL pricing of the regime‑switching BSM put option achieves very low out‑of‑sample errors (MAE ≈ 0.0092, MSE ≈ 0.0005). predictive-accuracy
    Source sentence
    Figure 2 (a) and (b) 0.0092 0.0005 0.0092 0.0012 Figure 2 (c) 0.0152 0.0006 0.0225 0.0013
  • For the regime‑switching Heston model, out‑of‑sample MAE/MSE range from 0.0199/0.0005 (ITM) to 0.0960/0.0135 (OTM). predictive-accuracy
    Source sentence
    Table 6: Out of sample prediction errors for the Heston stochastic volatility with regime switching put option prices for ITM, ATM and OTM option types.
  • PIRL networks produce option price paths that closely match Monte‑Carlo benchmarks across time to expiry for ITM, ATM and OTM strikes. pricing-calibration
    Source sentence
    Figure 4: European put option prices in the two regimes under Heston stochastic volatility model
  • The trained PIRL models are nearly instantaneous to evaluate once trained, eliminating the need for retraining when parameters change. performance
    Source sentence
    Specifically, we demonstrate that PIRLs eliminate the need for retraining and become nearly instantaneous once trained, thus, offering an efficient and flexible tool for pricing options across a broad range of specifications and parameters.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 34 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 (39)

  1. Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations (2019) — link
  2. Option valuation and hedging strategies with jumps in the volatility of asset returns (1993)
  3. Pricing variance and volatility swaps in a stochastic volatility model with regime switching: discrete observations case (2013)
  4. From characteristic function to distribution function: a simple framework for the theory (1991)
  5. Pricing volatility swaps under H eston’s stochastic volatility model with regime switching (2007)
  6. An analytical approximation formula for E uropean option pricing under a new stochastic volatility model with regime-switching (2016)
  7. A new exact solution for pricing E uropean options in a two-state regime-switching economy (2012)
  8. Regime switching and E uropean options (2002)
and 31 more
  1. Multilayer feedforward networks are universal approximators (1989) — link
  2. Deep residual learning for image recognition (2016)
  3. Algorithm 778: L-BFGS-B : F ortran subroutines for large-scale bound-constrained optimization (1997)
  4. Bankruptcy forecasting: An empirical comparison of A da B oost and neural networks (2008) — link
  5. A neural network approach for credit risk evaluation (2008) — link
  6. Natural language processing (2020)
  7. Forecasting Stock Indices: Stochastic and Artificial Neural Network Models (2024)
  8. Option pricing using Machine Learning (2021) — link
  9. Machine learning for quantitative finance: fast derivative pricing, hedging and fitting (2018) — link
  10. Machine learning for pricing A merican options in high-dimensional M arkovian and non- M arkovian models (2020) — link
  11. Predicting the stock price of frontier markets using machine learning and modified B lack– S choles Option pricing model (2020) — link
  12. D eep O ption: A novel option pricing framework based on deep learning with fused distilled data from multiple parametric methods (2021) — link
  13. hp- VPINNs : V ariational physics-informed neural networks with domain decomposition (2021)
  14. B-PINNs : B ayesian physics-informed neural networks for forward and inverse PDE problems with noisy data (2021)
  15. Piat: P hysics informed adversarial training for solving partial differential equations (2022) — link
  16. On the Limitations of Physics-Informed Deep Learning: Illustrations Using First-Order Hyperbolic Conservation Law-Based Traffic Flow Models (2023) — link
  17. Physics Informed Deep Learning for Traffic State Estimation (2020) — link
  18. Physics-Informed Deep Learning for Traffic State Estimation: A Hybrid Paradigm Informed By Second-Order Traffic Models (2021)
  19. A Physics-Informed Deep Learning Paradigm for Traffic State and Fundamental Diagram Estimation (2022) — link
  20. Empirical Asset Pricing via Ensemble G aussian Process Regression — → 2212.01048
  21. Accelerated A merican option pricing with deep neural networks (2023)
  22. Machine learning for quantitative finance: fast derivative pricing, hedging and fitting (2018)
  23. Empirical deep hedging (2023)
  24. Physics-Informed Deep Learning for Traffic State Estimation: Illustrations With LWR and CTM Models (2022) — link
  25. Option pricing in the H eston model with physics inspired neural networks (2024)
  26. Valuation of guaranteed minimum accumulation benefits ( GMAB s) with physics-inspired neural networks (2023)
  27. Option Pricing and Local Volatility Surface by Physics-Informed Neural Network (2024)
  28. A deep learning based numerical PDE method for option pricing (2023)
  29. Why R es N et Works? Residuals Generalize (2020) — link
  30. Power-enhanced residual network for function approximation and physics-informed inverse problems (2024) — link
  31. A physics-informed data-driven algorithm for ensemble forecast of complex turbulent systems (2024) — link

1 of 39 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.