Conditional Monte Carlo Learning for Diffusions II: extended methodology and application to risk measures and early stopping problems - Laboratoire de Probabilités et Modèles Aléatoires Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2020

Conditional Monte Carlo Learning for Diffusions II: extended methodology and application to risk measures and early stopping problems

Résumé

In Conditional Monte Carlo Learning for Diffusions part I (CMCLDI) [2], we presented a One-layered Nested Monte Carlo (1NMC) to simulate functionals U of a Markov process X. Based on a judicious combination between regression and 1NMC used for localization purpose, this methodology allows to simulate U_{t≥s} conditionally on X{s}. The parallel suitability and scalability of 1NMC makes this algorithm very competitive to simulate quantities that are almost impossible to simulate with other methods. In this paper, using the double layer of trajectories, we explain further the mathematical background of the control on the bias propagation. With this double layer structure, we also detail how to adjust the variance to get a better approximation of the second moment from the regression. In normal and log-normal models, this variance adjustment allows a better description of tail events. Since we applied this algorithm on Backward Stochastic Differential Equations in CMCLDI, we show here its strength for the simulation of risk measures and optimal stopping problems. Two highly dimensional numerical examples are executed in few minutes on one Graphics Processing Unit (GPU).
Fichier principal
Vignette du fichier
Cond.Learn II.pdf (1.49 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02959494 , version 1 (06-10-2020)

Identifiants

  • HAL Id : hal-02959494 , version 1

Citer

Lokman Abbas-Turki, G. Pagès, Babacar Diallo. Conditional Monte Carlo Learning for Diffusions II: extended methodology and application to risk measures and early stopping problems. 2020. ⟨hal-02959494⟩
822 Consultations
68 Téléchargements

Partager

Gmail Facebook X LinkedIn More