Embedologist

Library for experimentation with manifold learning methods

Juan M. Bello-Rivas

Introduction

Embedologist is a collection of routines for experimenting with manifold learning methods. It aims to be fast by relying on GPU backends (currently CuPy and JAX). This library is the successor of the diffusion-maps module.

Features

  • Dimensionality reduction via diffusion maps. Diffusion maps are differentiable: both with respect to the data set and with respect to out of distribution points.
  • Differentiable Gaussian process regression using GPU backends.

Installation

Follow the steps in INSTALL.md.

Support

The mailing list embedologist-devel is for development discussion and patches related to this project. For help sending patches to this list, please consult git-send-email.io.

S
Description
Library for experimentation with manifold learning methods
Readme MIT
350 KiB
Languages
Python 100%