geometric_ml

This repository contains:

• Matlab code (not maintained)
• Python code

to apply Riemannian geometry in machine learning.

1. Construct Riemannian metrics.
2. Compute shortest paths.
3. Fit the LAND model.
4. Change the geometry of the ambient space.

Most of the methods have been proposed in the following papers:

• "A Locally Adaptive Normal Distribution", G. Arvanitidis, et. al., NeurIPS 2016
• "Latent Space Oddity: on the Curvature of Deep Generative Models", G. Arvanitidis, et. at., ICLR 2018
• "Fast and Robust Shortest Paths on Data Learned Manifolds", G. Arvanitidis, et. al., AISTATS 2019
• "Geometrically Enriched Latent Spaces", G. Arvanitidis, et. al., AISTATS 2021

This code is published for research purposes only. Please cite the corresponding papers in case you use parts of this code base.