Welcome to FisherEyes

The fishereyes package provides tools to learn smooth, invertible transformations of data where each point represents a measurement and is associated with its own uncertainty in the form of a covariance matrix. It enables the transformation of locally anisotropic data into a space where push-forward uncertainties are isotropic and uniform. The core FisherEyes class offers a modular interface to plug in different transformation models, loss functions, and optimizers – supporting uncertainty-aware learning in a wide range of scientific and machine learning tasks.

Installation

The Python package fishereyes can be installed from PyPI:

python -m pip install fishereyes

Development installation

If you want to contribute to the development of fishereyes, we recommend the following editable installation from this repository:

git clone https://github.com/william-h-oliver/fishereyes
cd fishereyes
python -m pip install --editable .[tests]

Having done so, the test suite can be run using pytest:

python -m pytest

Acknowledgments

This repository was set up using the SSC Cookiecutter for Python Packages.