Software

Packages I developed

I have developed Python packages implementing some of my work on statistical models and algorithms. These packages use a PyTorch backend, they are well documented, and use best practices for software development such as testing and continuous integration.

• sqfa: A Python package for learning linear features that maximize second-order differences between classes, using information geometry.
• projnormal: A Python package for working with the projected normal distribution in the N-dimensional sphere.

Packages I contribute to

I have contributed to the following packages:

• scikit-learn: Python package for machine learning. I incorporated robust covariance estimation into Quadratic Discriminant Analysis (#32108) and fixed a bias in the Minimum Covariance Determinant estimator (#32117
• geomstats: A Python package for performing computations and statistical analyses on manifolds. I improved the computation of distances between SPD matrices.
• plenoptic: A Python package for model-based synthesis of perceptual stimuli. I contributed to refactoring the texture synthesis model.