AReS-MaRS

This repository contains code that implements the AReS and MaRS algorithms as presented in the paper:

AReS and MaRS - Adversarial and MMD-Minimizing Regression for SDEs. Gabriele Abbati*, Philippe Wenk*, Michael A Osborne, Andreas Krause, Bernhard Schölkopf and Stefan Bauer. In Proceedings of the 36th International Conference on Machine Learning (ICML), 2019.
*: equal contribution

Code

The code provided is written in Python 3.6, and relies on the following libraries:

• TensorFlow
• numpy
• scipy

Some usage examples (the same ones used in the experimental section of the paper) can be found in the examples/ directory.