Nacira Agram: SIG-BSDE for Dynamic Risk Measures

V četrtek, 8. 5. 2025, ob 14:15 bo v predavalnici 3.07 v okviru seminarja VeSFiM potekalo predavanje Nacire Agram (KTH University, Stockholm, Švedska) z naslovom SIG-BSDE for Dynamic Risk Measures.

Povzetek: In this talk, we introduce SIG-BSDE, a novel algorithm for numerically solving backward stochastic differential equations (BSDEs). The method integrates path signatures with the backward Euler–Maruyama scheme, resulting in a robust and efficient numerical solver. We present a rigorous convergence analysis of SIG-BSDE and validate its effectiveness through a series of numerical experiments, benchmarking its performance against existing machine learning-based methods. Our primary application focus is on BSDEs arising in the context of dynamic risk measures. In this setting, we also propose a new framework for modeling dynamic risk under generalized ambiguity. We demonstrate how a hybrid approach, combining deep learning techniques with the SIG-BSDE algorithm, can effectively tackle such problems and offer a promising direction for risk-aware decision-making under uncertainty.

Predavanje bo potekalo v živo.

Vljudno vabljeni!