Code
Code accompanying my research papers and numerical experiments can be found on GitHub .
Stochastic control with expected signatures
Implementation of the expected signature formula via coordinate change. An illustration solving the optimal tracking of fractional Brownian motion is presented.
Time-series learning with Volterra signatures
Regression and classification experiments with real-world data and time-series features built from the Volterra signature. The code relies on the JAX-based Python package tensordev, see tensordev, which implements algorithms from the accompanying technical paper on computational aspects of the Volterra signature.
Rough PDEs for European options and Greeks
Code for pricing European options and computing Greeks in general local stochastic volatility models via a finite-difference implementation of the rough PDE pricing method.
Optimal stopping with signatures
Implementation of primal and dual algorithms for solving non-Markovian optimal stopping problems with path signatures. The examples include stopping problems for fractional Brownian motion, American option pricing under rough-volatility models such as rough Bergomi and rough Heston, and extensions to deep and kernel learning.
Local regression with signatures
Local kernel regression and classification on path space using signature-based distances. The repository contains implementations for path-valued regression and classification tasks, including examples for time series, SDE learning, and path-dependent option pricing.