Seminar: "Stability of Backward Propagation of Chaos and Numerical Approaches"

ΚΥΚΛΟΣ ΣΕΜΙΝΑΡΙΩΝ ΣΤΑΤΙΣΤΙΚΗΣ 2025-2026

Presenter: Alexandros Saplaouras, Post-Doctoral Researcher, Insurance Mathematics and Stochastic Finance, Department of Mathematics, ETH Zurich

Room: 709, Evelpidon Building

ABSTRACT

We study the asymptotic behavior of mean-field systems of backward stochastic differential equations with jumps (BSDEs). As the number of interacting equations tends to infinity, their solutions converge to independent and identically distributed copies of the limiting McKean–Vlasov BSDE (MVBSDE), a phenomenon known as backward propagation of chaos. We establish a general framework ensuring that this property is preserved, in both discrete- and continuous time settings. Finally, we discuss numerical approximation approaches for (MV-)BSDEs based on neural network techniques, and we present some results for specific cases.