Optimal stopping with signatures (Q6103968)

The objective of this paper is to propose a new method for solving optimal stopping problems (such as American option pricing in finance) under minimal assumptions on the underlying stochastic process \(X\). The decision-makers strategies are stopping times (and randomized stopping times) represented by linear and nonlinear functional of the rough path signature \(\mathbb{X}^{<\infty}\) associated to process \(X\) (cf. \textit{B. Hambly} and \textit{T. Lyons} [Ann. Math. (2) 171, No. 1, 109--167 (2010; Zbl 1276.58012)]), and prove that maximizing over these classes of signature stopping times, solves the original optimal stopping problem. Using the algebraic properties of the signature, the problem is reformulated to a (deterministic) optimization problem depending only on the (truncated) expected signature \(\mathbf{E}[\mathbb{X}^{\leq N}_{0,T} ]\). Next, applying a deep neural network approach to approximate the nonlinear signature functional, it is efficiently solved numerically. The only assumption on the process \(X\) is that it is a continuous (geometric) random rough path. The theory encompasses processes such as fractional Brownian motion, which fail to be either semimartingales or Markov processes, and can be used, in particular, for American-type option pricing in fractional models, for example, on financial or electricity markets.