Optimal Stopping with Adapted Neural Networks
8 Pages Posted: 12 Dec 2019 Last revised: 13 Dec 2019
Date Written: November 22, 2019
Abstract
We use temporally adapted neural networks to solve a generalization of the optimal exercise problem for a discrete set of possible exercise times. Versions based on convolutional and attention layers were implemented, tested and found to produce state of the art results on the fractional Brownian motion with various Hurst parameters. The approach is intuitive and fully agnostic with respect to the dependency structure of the underlying stochastic process.
Keywords: optimal stopping, optimal exercise, american options, deep neural networks, convolutional networks, attention networks
JEL Classification: C41, C45, G13
Suggested Citation: Suggested Citation