ARTEMIS: A Neuro Symbolic Framework for Economically Constrained Market Dynamics
arXiv:2603.18107v1 Announce Type: new 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 ARTEM
arXiv:2603.18107v1 Announce Type: new 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.
Executive Summary
This article presents ARTEMIS, a novel neuro-symbolic framework that addresses the limitations of deep learning models in quantitative finance by incorporating fundamental economic principles such as no-arbitrage constraints. ARTEMIS combines a Laplace Neural Operator encoder, a neural stochastic differential equation, and a differentiable symbolic bottleneck to produce interpretable trading rules. The model is evaluated against six strong baselines on four datasets and achieves state-of-the-art directional accuracy. A comprehensive ablation study confirms the contribution of each component. The results demonstrate the potential of ARTEMIS to bridge the gap between deep learning's power and the transparency demanded in quantitative finance.
Key Points
- ▸ ARTEMIS is a neuro-symbolic framework that combines a Laplace Neural Operator encoder, a neural stochastic differential equation, and a differentiable symbolic bottleneck.
- ▸ The model enforces economic plausibility via two novel regularisation terms: a Feynman-Kac PDE residual and a market price of risk penalty.
- ▸ ARTEMIS achieves state-of-the-art directional accuracy on four datasets and outperforms all baselines on two datasets.
Merits
Strength in Interpretability
ARTEMIS produces interpretable trading rules, addressing a key limitation of deep learning models in quantitative finance.
Incorporation of Economic Principles
The model enforces economic plausibility via novel regularisation terms, ensuring that predictions are grounded in fundamental economic principles.
Demerits
Potential Complexity
The model's architecture may be complex, potentially making it challenging to implement and interpret.
Dataset Requirements
ARTEMIS may require large and diverse datasets to effectively train and evaluate its performance.
Expert Commentary
ARTEMIS represents a significant advancement in the field of quantitative finance, addressing the limitations of deep learning models by incorporating fundamental economic principles. The model's interpretability features and state-of-the-art performance on various datasets make it an attractive solution for practitioners and regulators alike. However, the potential complexity of the model and the requirement for large and diverse datasets may limit its adoption. Furthermore, the development of ARTEMIS highlights the need for regulatory frameworks that support the use of advanced AI models in finance, while ensuring transparency and accountability.
Recommendations
- ✓ Future research should focus on scaling up ARTEMIS to handle larger datasets and more complex market scenarios.
- ✓ The development of regulatory frameworks that support the use of advanced AI models in finance should prioritize transparency, accountability, and explainability.
