Conditional Expectation with Numerical Guarantees

Are you tired of black-box machine learning algorithms that lack transparency and fail to provide trustworthy predictions? Look no further!

This repository presents a novel approach developed by Prof. Patrick Cheridito and myself for computing conditional expectations with numerical guarantees.

Using an alternative expected value representation of the minimal L2 distance between Y and f(X) over all Borel measurable functions f, we provide guarantees for the accuracy of any numerical approximation of a given conditional expectation. We illustrate the method by assessing the quality of numerical approximations to different high-dimensional nonlinear regression problems.

For more details on the theoretical background and methodology used in this library, please refer to the following research paper:

Computation of conditional expectations with guarantees
arXiv Preprint 2112.01804
Journal of Scientific Computing 95(12), 2023 Springer

Table of Contents

• Installation
• Usage
• Citation

Running Code

Installation

Start by installing the necessary dependencies. You can do this by running the following command:

pip install -r requirements.txt

Usage

To execute the experiment, use the following command-line input:

python3 linear_example.py

Citation

If you use this work in your research or publication, please cite the following paper:

@article{cheridito2023computation,
  title={Computation of conditional expectations with guarantees},
  author={Cheridito, Patrick and Gersey, Balint},
  journal={Journal of Scientific Computing},
  volume={95},
  number={1},
  pages={12},
  year={2023},
  publisher={Springer}
}