Search papers, labs, and topics across Lattice.
This paper introduces two neural-network-based numerical schemes for solving systems of coupled ergodic Backward Stochastic Differential Equations (eBSDEs), aimed at optimizing strategies in regime-switching stochastic factor models. By establishing a connection between eBSDE solutions and a multidimensional BSDE with random terminal time, the authors develop a locally additive deep learning scheme and a new Deep Galerkin Method (DGM) to minimize residuals in the associated ergodic PDE system. Numerical experiments validate the effectiveness of these methods, highlighting the significant influence of regime switches on forward utility preferences.
Neural network techniques can effectively solve complex systems of ergodic BSDEs, revealing how regime switches dramatically alter forward utility preferences.
In this paper, we introduce two neural-network-based numerical schemes for solving systems of coupled ergodic Backward Stochastic Differential Equations (eBSDEs), motivated by the approximation of optimal strategies within the framework of forward utilities in a regime-switching stochastic factor model. Our approach builds on the representation of such models through systems of eBSDEs introduced in [HLT20]. We first establish a link between the solution of the system of ergodic BSDEs and that of an associated multidimensional BSDE with random terminal time, given by the hitting time of the positive recurrent stochastic factor. Building on this representation, we introduce a locally additive deep learning scheme obtained by minimizing aggregated local error terms. We then present a new Deep Galerkin Method (DGM) inspired algorithm that minimizes the residual of the associated ergodic PDE system, relying on a representation of the ergodic cost. Finally, we apply this framework to regime-switching forward utilities in a stochastic factor model. We first derive a general consistency SPDE that characterizes regime-switching forward utilities and retrieve their representation with systems of ergodic BSDEs in the homothetic case. Numerical experiments demonstrate the performance of the proposed methods, with a particular focus on the impact on forward preferences of taking into account regime switches.
