Skip to content

A high performance topological machine learning toolbox in Python

License

Notifications You must be signed in to change notification settings

moneytech/giotto-tda

 
 

Repository files navigation

image

Version_ _ Azure-cov_ Azure-test_ Twitter-follow_ Slack-join_

giotto-tda

giotto-tda is a high performance topological machine learning toolbox in Python built on top of scikit-learn and is distributed under the GNU AGPLv3 license. It is part of the Giotto family of open-source projects.

Project genesis

giotto-tda is the result of a collaborative effort between L2F SA, the Laboratory for Topology and Neuroscience at EPFL, and the Institute of Reconfigurable & Embedded Digital Systems (REDS) of HEIG-VD.

Documentation

Getting started

To get started with giotto-tda, first follow the installations steps below. This blog post, and references therein, offer a friendly introduction to the topic of topological machine learning and to the philosophy behind giotto-tda.

Tutorials and use cases

Simple tutorials can be found in the examples folder. For a wide selection of use cases and application domains, you can visit this page.

Installation

Dependencies

The latest stable version of giotto-tda requires:

  • Python (>= 3.5)
  • NumPy (>= 1.17.0)
  • SciPy (>= 0.17.0)
  • joblib (>= 0.11)
  • scikit-learn (>= 0.22.0)
  • python-igraph (>= 0.7.1.post6)
  • matplotlib (>= 3.0.3)
  • plotly (>= 4.4.1)
  • ipywidgets (>= 7.5.1)

To run the examples, jupyter is required.

User installation

Linux and macOS

The simplest way to install giotto-tda is using pip :

pip install -U giotto-tda

If necessary, this will also automatically install all the above dependencies. Note: we recommend upgrading pip to a recent version as the above may fail on very old versions.

Pre-release, experimental builds containing recently added features, and/or bug fixes can be installed by running :

pip install -U giotto-tda-nightly

The main difference between giotto-tda-nightly and the developer installation (see below) is that the former is shipped with pre-compiled wheels (similarly to the stable release) and hence does not require any C++ dependencies.

Windows

In this case, python-igraph and its dependency pycairo must be manually installed before proceeding as above. This is because the python-igraph project does not yet provide official installers for Windows via PyPI, so that pip install python-igraph would fail there. The preferred way to install python-igraph on Windows is to download and install the relevant wheels built by Christoph Gohlke for both pycairo and python-igraph. We host these wheels so they can be fetched with convenient URLs. For Python 3.5 to 3.7, you may run :

pip install https://storage.googleapis.com/l2f-open-models/giotto-learn/windows-binaries/pycairo/pycairo-1.18.2-cp<PYTHON VERSION>-cp<PYTHON VERSION>m-win_amd64.whl
pip install https://storage.googleapis.com/l2f-open-models/giotto-learn/windows-binaries/python-igraph/python_igraph-0.7.1.post6-cp<PYTHON VERSION>-cp<PYTHON VERSION>m-win_amd64.whl

where <PYTHON VERSION> is e.g. 37 for Python 3.7. For Python 3.8, you may run :

pip install https://storage.googleapis.com/l2f-open-models/giotto-learn/windows-binaries/pycairo/pycairo-1.18.2-cp38-cp38-win_amd64.whl
pip install https://storage.googleapis.com/l2f-open-models/giotto-learn/windows-binaries/python-igraph/python_igraph-0.7.1.post6-cp38-cp38-win_amd64.whl

Contributing

We welcome new contributors of all experience levels. The Giotto community goals are to be helpful, welcoming, and effective. To learn more about making a contribution to giotto-tda, please see the CONTRIBUTING.rst file.

Developer installation

Installing both the PyPI release and source of giotto-tda in the same environment is not recommended since it is known to cause conflicts with the C++ bindings. On Windows, the pycairo and python-igraph dependencies have to be installed manually just as in the case of a simple user installation.

The developer installation requires three important C++ dependencies:

  • A C++14 compatible compiler
  • CMake >= 3.9
  • Boost >= 1.56

Please refer to your system's instructions and to the CMake and Boost websites for definitive guidance on how to install these dependencies. The instructions below are unofficial, please follow them at your own risk.

Linux

Most Linux systems should come with a suitable compiler pre-installed. For the other two dependencies, you may consider using your distribution's package manager, e.g. by running

sudo apt-get install cmake boost

if apt-get is available in your system.

macOS

On macOS, you may consider using brew (https://brew.sh/) to install the dependencies as follows:

brew install gcc cmake boost

Windows

On Windows, you will likely need to have Visual Studio installed. At present, it appears to be important to have a recent version of the VS C++ compiler. One way to check whether this is the case is as follows: 1) open the VS Installer GUI; 2) under the "Installed" tab, click on "Modify" in the relevant VS version; 3) in the newly opened window, select "Individual components" and ensure that v14.24 or above of the MSVC "C++ x64/x86 build tools" is selected. The CMake and Boost dependencies are best installed using the latest binary executables from the websites of the respective projects.

Source code

You can obtain the latest state of the source code with the command:

git clone https://github.com/giotto-ai/giotto-tda.git

To install:

cd giotto-tda
pip install -e ".[tests, doc]"

This way, you can pull the library's latest changes and make them immediately available on your machine. Note: we recommend upgrading pip and setuptools to recent versions before installing in this way.

Testing

After installation, you can launch the test suite from outside the source directory:

pytest gtda

Changelog

See the RELEASE.rst file for a history of notable changes to giotto-tda.

Community

giotto-ai Slack workspace: https://slack.giotto.ai/

Contacts

maintainers@giotto.ai

About

A high performance topological machine learning toolbox in Python

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C++ 70.9%
  • Python 27.4%
  • Other 1.7%