DeepMartingale: Duality of the Optimal Stopping Problem with Expressivity

2025/10/13 by Ye, Junyan, Wong, Hoi Ying
#FOS: Computer and information sciences #FOS: Mathematics #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Numerical Analysis (math.NA) #Optimization and Control (math.OC) #Probability (math.PR)

paper · doi:10.48550/arxiv.2510.13868

Abstract

Using a martingale representation, we introduce a novel deep-learning approach, which we call DeepMartingale, to study the duality of discrete-monitoring optimal stopping problems in continuous time. This approach provides a tight upper bound for the primal value function, even in high-dimensional settings. We prove that the upper bound derived from DeepMartingale converges under very mild assumptions. Even more importantly, we establish the expressivity of DeepMartingale: it approximates the true value function within any prescribed accuracy ε under our architectural design of neural networks whose size is bounded by c D^qε^-r, where the constants c, q, r are independent of the dimension D and the accuracy ε. This guarantees that DeepMartingale does not suffer from the curse of dimensionality. Numerical experiments demonstrate the practical effectiveness of DeepMartingale, confirming its convergence, expressivity, and stability.

Citations

• A deep primal-dual BSDE method for optimal stopping problems
• A pure dual approach for hedging Bermudan options
• Neural reproducing kernel Banach spaces and representer theorems for deep networks
• Deep neural network expressivity for optimal stopping problems
• Neural Optimal Stopping Boundary
• Understanding neural networks with reproducing kernel Banach spaces
• Optimal Stopping via Randomized Neural Networks
• Deep neural network approximation for high-dimensional parabolic Hamilton-Jacobi-Bellman equations
• Approximation Bounds for Random Neural Networks and Reservoir Systems
• Approximation bounds for random neural networks and reservoir systems
• Neural network regression for Bermudan option pricing
• Simulation-based Value-at-Risk for Nonlinear Portfolios
• Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations
• Mean Field Forward-Backward Stochastic Differential Equations
• Optimal dual martingales, their analysis and application to new algorithms for Bermudan products

Related

• Deep neural network expressivity for optimal stopping problems
• Deep Learning for the Multiple Optimal Stopping Problem
• Deep Penalty Methods: A Class of Deep Learning Algorithms for Solving High Dimensional Optimal Stopping Problems
• Optimal Stopping with Trees: The Details
• A deep primal-dual BSDE method for optimal stopping problems