Deep BSVIEs Parametrization and Learning-Based Applications

TL;DR

This paper develops a new numerical approximation method for complex backward stochastic Volterra integral equations using deep learning, enabling applications in areas like recursive utilities with memory.

Contribution

It introduces a well-posedness framework for BSVIEs and extends deep neural network-based solvers to handle their two-dimensional time structure and reflected variants.

Findings

Proposed a convergent deep learning scheme for BSVIEs.

Extended the solver to reflected BSVIEs with constraints.

Provided a rigorous convergence analysis for the method.

Abstract

We study the numerical approximation of backward stochastic Volterra integral equations (BSVIEs) and their reflected extensions, which naturally arise in problems with time inconsistency, path dependent preferences, and recursive utilities with memory. These equations generalize classical BSDEs by involving two dimensional time structures and more intricate dependencies. We begin by developing a well posedness and measurability framework for BSVIEs in product probability spaces. Our approach relies on a representation of the solution as a parametrized family of backward stochastic equations indexed by the initial time, and draws on results of Stricker and Yor to ensure that the two parameter solution is well defined in a joint measurable sense. We then introduce a discrete time learning scheme based on a recursive backward representation of the BSVIE, combining the discretization of Hamaguchi and Taguchi with deep neural networks. A detailed convergence analysis is provided, generalizing the framework of deep BSDE solvers to the two dimensional BSVIE setting. Finally, we extend the solver to reflected BSVIEs, motivated by applications in delayed recursive utility with lower constraints.