Paper: arXiv 2603.18107

Authors: Rahul D Ray

Abstract

Deep learning models in quantitative finance often operate as black boxes, lacking interpretability and failing to incorporate fundamental economic principles such as no-arbitrage constraints. This paper introduces ARTEMIS (Arbitrage-free Representation Through Economic Models and Interpretable Symbolics), a novel neuro-symbolic framework combining a continuous-time Laplace Neural Operator encoder, a neural stochastic differential equation regularised by physics-informed losses, and a differentiable symbolic bottleneck that distils interpretable trading rules. The model enforces economic plausibility via two novel regularisation terms: a Feynman-Kac PDE residual penalising local no-arbitrage violations, and a market price of risk penalty bounding the instantaneous Sharpe ratio. We evaluate ARTEMIS against six strong baselines on four datasets: Jane Street, Optiver, Time-IMM, and DSLOB (a synthetic crash regime). Results demonstrate ARTEMIS achieves state-of-the-art directional accuracy, outperforming all baselines on DSLOB (64.96%) and Time-IMM (96.0%). A comprehensive ablation study confirms each component’s contribution: removing the PDE loss reduces directional accuracy from 64.89% to 50.32%. Underperformance on Optiver is attributed to its long sequence length and volatility-focused target. By providing interpretable, economically grounded predictions, ARTEMIS bridges the gap between deep learning’s power and the transparency demanded in quantitative finance.

Complexity vs Empirical Score

  • Math Complexity: 8.5/10
  • Empirical Rigor: 8.0/10
  • Quadrant: Holy Grail — high math complexity, high empirical rigor

Why this score: ARTEMIS presents an innovative neuro-symbolic framework incorporating Neural SDEs, Laplace Neural Operators, and physics-informed losses to enforce hard economic constraints like no-arbitrage. The empirical validation is strong across multiple datasets, including stress-testing on extreme market regimes, backed by rigorous ablation studies.

Research Flowchart

  flowchart TD
    A[Goal: Build an interpretable, arbitrage-free deep learning model for market dynamics] --> B[Inputs: Jane Street, Optiver, Time-IMM, & Synthetic Crash DSLOB Datasets]
    B --> C[ARTEMIS Architecture: Neural Operator Encoder + Neural SDE Driven Bottleneck]
    C --> D[Economic Physics-Informed Regularization: Feynman-Kac PDE & Sharpe Ratio Penalties]
    D --> E[Differentiable Symbolic Bottleneck: Distills Interpretable Trading Rules]
    E --> F[Outcomes: SOTA Accuracy on DSLOB 64.96% & Time-IMM 96.0%; PDE Loss Critical for Performance]

Cited References (58)

  1. Deep learning for financial time series prediction: A state-of-the-art review of standalone and hybrid models (2024)
  2. Deep learning models for price forecasting of financial time series: A review of recent advancements: 2020–2022 (2024)
  3. Financial time series forecasting: A comprehensive review of signal processing and optimization-driven intelligent models (2025)
  4. Capability of Hybrid Long Short-Term Memory in Stock Price Prediction: A Comprehensive Literature Review (2024)
  5. Deep learning for financial forecasting: A review of recent trends (2025)
  6. Machine learning for quantitative finance applications: A survey (2019)
  7. Machine learning for financial risk management: a survey (2020)
  8. Machine learning algorithms for financial asset price forecasting (2020) — link
and 50 more
  1. An overview of machine learning, deep learning, and reinforcement learning-based techniques in quantitative finance: recent progress and challenges (2023)
  2. Explainable Artificial Intelligence (XAI) approaches for transparency and accountability in financial decision-making (2023)
  3. Interpretable trading pattern designed for machine learning applications (2023)
  4. Explainable AI in Finance: Enhancing Transparency and Interpretability of AI Models in Financial Decision-Making (2026)
  5. Recent emerging techniques in explainable artificial intelligence to enhance the interpretable and understanding of AI models for human (2025)
  6. Artificial Intelligence in Financial Trading Predictive Models and Risk Management Strategies (2025)
  7. Machine learning for financial prediction under regime change using technical analysis: A systematic review (2023)
  8. Predictive Modeling of US Stock Market Trends Using Hybrid Deep Learning and Economic Indicators to Strengthen National Financial Resilience (2023)
  9. Time series forecasting model for non-stationary series pattern extraction using deep learning and GARCH modeling (2024)
  10. Deep learning for solving economic models (2025)
  11. The application of improved physics-informed neural network (IPINN) method in finance (2022)
  12. Physics-Informed Neural Networks for Option Pricing and Hedging in Illiquid Jump Markets (2025)
  13. Solving nonlinear and high-dimensional partial differential equations via deep learning (2018) — link
  14. Differential learning methods for solving fully nonlinear PDEs (2023)
  15. A Physics informed neural network approach for solving time fractional Black-Scholes partial differential equations: SM Nuugulu et al. (2025)
  16. Option pricing in the Heston model with physics inspired neural networks (2024)
  17. QUANTODE: A NEURAL DIFFERENTIAL EQUATION-BASED FRAMEWORK FOR CONTINUOUS-TIME FINANCIAL MARKET MODELING (2025)
  18. Risk and volatility: Econometric models and financial practice
  19. Forecasting developed and BRICS stock markets with cryptocurrencies and gold: generalized orthogonal generalized autoregressive conditional heteroskedasticity and generalized autoregressive score analysis (2022)
  20. Conditional heteroskedasticity in asset returns: A new approach (1991)
  21. Modeling long memory in stock market volatility (2000)
  22. Forecasting volatility by using wavelet transform, ARIMA and GARCH models (2023)
  23. Comparative analysis of ARIMA and GARCH models for forecasting spot gold prices and their volatility: a time series study (2023)
  24. Capturing long-memory properties in road fatality rate series by an autoregressive fractionally integrated moving average model with generalized autoregressive conditional heteroscedasticity: A case study of Florida, the United States, 1975–2018 (2022)
  25. On fractionally integrated autoregressive moving-average time series models with conditional heteroscedasticity (1997)
  26. Forecasting the volatilities of the Nigeria stock market prices (2017)
  27. Transformers for Multivariate Time Series Forecasting: Comprehensive Analysis, Challenges, Research Opportunities and Future Prospects (2026)
  28. Deep learning models for time series forecasting: A review (2024)
  29. Multivariate time series forecasting: A review (2024)
  30. LSTM vs. Transformer Models in Power Forecasting: A Comprehensive Survey (2026)
  31. Deep learning for time series forecasting: a survey (2025)
  32. A comprehensive survey of deep learning for time series forecasting: architectural diversity and open challenges (2025)
  33. Stock market index prediction using deep Transformer model (2022)
  34. The Time Series Informer Model for Stock Market Prediction (2025)
  35. Stock price prediction based on non-stationary transformers model (2023)
  36. Time series forecasting using transformer neural network (2024)
  37. Financial time series forecasting using CNN and transformer — → 2304.04912
  38. Stock market index prediction using transformer neural network models and frequency decomposition (2024)
  39. Comparison between transformer, informer, autoformer and non-stationary transformer in financial market (2023)
  40. Stock prediction method based on multi-source data of social media and Informer model (2025)
  41. Foundation models for time series analysis: A tutorial and survey (2024)
  42. Comprehensive review of neural differential equations for time series analysis (2025) — link
  43. Physics-informed neural network (PINN) evolution and beyond: A systematic literature review and bibliometric analysis (2022)
  44. Conformal prediction: A data perspective (2025)
  45. An overview on machine learning methods for partial differential equations: from physics informed neural networks to deep operator learning (2024) — link
  46. Empowering Time Series Analysis with Foundation Models: A Comprehensive Survey (2024) — link
  47. Neural Differential Equations for Continuous-Time Analysis (2025)
  48. Optiver Realized Volatility Prediction (2021)
  49. Jane Street Real-Time Market Data Forecasting (2024)
  50. Time-IMM: A Dataset and Benchmark for Irregular Multimodal Multivariate Time Series (2025) — link

1 of 58 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2518 linked papers.