TopologicalLearningTechniques 0.2.1

Package created for INF367AII at UiB

README

This package contains a small library of topological machine learning techniques. This package is made as part of a project for the Topological Machine Learning course at UiB.

Contents

This algorithms implemented in this package can be grouped into 3 groups:

• Graph fitting
  • Self organising maps
  • Growing Neural Gas
  • Reeb graphs
• Persistent Homology
  • The Rips-complex algorithm
  • Column reduction for persistence diagrams
  • Persistence Images
  • Persistence Landscapes
• Neural Networks
  • PersLay
  • Topological Autoencoder (WIP: the loss function doesn't back-propagate properly yet.)

Documentation

The documentation for this package is hosted on readthedocs.

Installing

This package imports Pytorch and therefore requires a Python version that supports pytorch: it has been developed and tested using Python 3.8.

To install this package simply use the following pip install bash command:

$ pip install TopologicalMachineLearningTechniques

When developing this package it is a good idea to install the package with the dev extras enabled. However, currently there are no additional dev requirements. You can install with dev dependencies using the following command:

$ pip install TopologicalMachineLearningTechniques[dev]

Usage

A few Jupyter notebooks have been created to demonstrate usage of the package:

• Graph Fitting notebook
• Persistent Homology notebook
• Neural Network notebook

These notebooks can be found on the GitHub page of this package.