Deep numerical schemes for systems of Ergodic BSDEs with applications to regime-switching forward utilities

2026-06-23 · Source: Takara TLDR - Daily AI Papers · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Quantitative Finance · Depth: Expert, quick

Summary

This paper introduces two novel neural-network-based numerical schemes designed to solve systems of coupled ergodic Backward Stochastic Differential Equations (eBSDEs). These schemes are developed to approximate optimal strategies within the context of forward utilities in a regime-switching stochastic factor model, building upon prior eBSDE representations. The first method is a locally additive deep learning scheme that minimizes aggregated local error terms, established by linking eBSDE solutions to a multidimensional BSDE with a random terminal time. The second is a Deep Galerkin Method (DGM)-inspired algorithm, which minimizes the residual of an associated ergodic PDE system using an ergodic cost representation. The framework is applied to regime-switching forward utilities, deriving a general consistency SPDE and demonstrating the methods' performance through numerical experiments, particularly highlighting the impact of regime switches on forward preferences.

Key takeaway

For research scientists developing optimal strategies in financial models with regime-switching dynamics, these deep numerical schemes offer robust methods for solving complex eBSDE systems. You should consider integrating these neural-network-based approaches, particularly the locally additive deep learning or DGM-inspired algorithms, to more accurately capture the impact of regime switches on forward preferences. This can lead to more precise approximations of optimal strategies and better model performance.

Key insights

Neural network schemes effectively solve coupled ergodic BSDEs for regime-switching forward utilities.

Principles

• eBSDE systems can represent optimal strategies in regime-switching models.
• Minimizing local error terms improves deep learning scheme accuracy.

Method

The approach links eBSDE solutions to a multidimensional BSDE with random terminal time, then applies either a locally additive deep learning scheme or a DGM-inspired algorithm minimizing PDE residuals.

In practice

• Implement locally additive deep learning for eBSDEs.
• Adapt DGM for ergodic PDE system residuals.

Topics

• Ergodic BSDEs
• Deep Learning Schemes
• Regime-Switching Models
• Forward Utilities
• Stochastic Factor Models
• Deep Galerkin Method

Best for: AI Scientist, Research Scientist

Related on AIssential

• Stochastic Penalty-Barrier Methods for Constrained Machine Learning — SPBM offers a general method for constrained machine learning in non-convex, non-smooth, stochastic deep learning settings.
• A Bifurcation Theory Framework for Gradient Descent on the Edge of Stability — A bifurcation theory framework explains stable Edge of Stability training in overparameterized neural networks via flip bifurcations and tangent dynamics.
• Deep regression learning from dependent observations with minimum error entropy principle — MEE-based deep neural networks achieve minimax optimal rates for nonparametric regression with dependent data.
• Stochastic Gradient Methods: Bias, Stability and Generalization — A new framework analyzes stability and generalization for biased stochastic gradient methods.
• Deep Learning Surrogates for Emulating Stochastic Climate Tipping Dynamics — A dynamics-informed Temporal Fusion Transformer provides a 465x faster, differentiable surrogate for forecasting stochastic climate tipping events with high fidelity.
• Distributionally Robust Geometric Joint Chance-Constrained Optimization: Neurodynamic Approaches — A neurodynamic duplex neural network solves distributionally robust chance-constrained optimization problems, converging to global optima.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.