FEDOT concepts

December 26, 2022 · View on GitHub

.. |eng| image:: https://img.shields.io/badge/lang-en-red.svg :target: /README_en.rst

.. |rus| image:: https://img.shields.io/badge/lang-ru-yellow.svg :target: /README.rst

.. image:: /docs/fedot_logo.png :alt: Logo of FEDOT framework

.. start-badges .. list-table:: :stub-columns: 1

    • package
    • | |pypi| |python|
    • tests
    • | |build| |integration| |coverage|
    • docs
    • |docs|
    • license
    • | |license|
    • stats
    • | |downloads_stats|
    • support
    • | |tg|
    • languages
    • | |eng| |rus|
    • mirror
    • | |gitlab|
    • funding
    • | |ITMO| |NCCR| .. end-badges

FEDOT is an open-source framework for automated modeling and machine learning (AutoML) problems. This framework is distributed under the 3-Clause BSD license.

It provides automatic generative design of machine learning pipelines for various real-world problems. The core of FEDOT is based on an evolutionary approach and supports classification (binary and multiclass), regression, clustering, and time series prediction problems.

.. image:: /docs/fedot-workflow.png :alt: The structure of the AutoML workflow in FEDOT

The key feature of the framework is the complex management of interactions between various blocks of pipelines. It is represented as a graph that defines connections between data preprocessing and model blocks.

The project is maintained by the research team of the Natural Systems Simulation Lab, which is a part of the National Center for Cognitive Research of ITMO University <https://actcognitive.org/>__.

More details about FEDOT are available in the next video:

.. image:: https://res.cloudinary.com/marcomontalbano/image/upload/v1606396758/video_to_markdown/images/youtube--RjbuV6i6de4-c05b58ac6eb4c4700831b2b3070cd403.jpg :target: http://www.youtube.com/watch?v=RjbuV6i6de4 :alt: Introducing Fedot

FEDOT concepts

  • Flexibility. FEDOT can be used to automate the construction of solutions for various problems <https://fedot.readthedocs.io/en/master/introduction/fedot_features/main_features.html#involved-tasks>, data types <https://fedot.readthedocs.io/en/master/introduction/fedot_features/automation_features.html#data-nature> (texts, images, tables), and models <https://fedot.readthedocs.io/en/master/advanced/automated_pipelines_design.html>_;
  • Extensibility. Pipeline optimization algorithms are data- and task-independent, yet you can use special strategies <https://fedot.readthedocs.io/en/master/api/strategies.html>_ for specific tasks or data types (time-series forecasting, NLP, tabular data, etc.) to increase the efficiency;
  • Integrability. FEDOT supports widely used ML libraries (Scikit-learn, CatBoost, XGBoost, etc.) and allows you to integrate custom ones <https://fedot.readthedocs.io/en/master/api/strategies.html#module-fedot.core.operations.evaluation.custom>_;
  • Tuningability. Various hyper-parameters tuning methods <https://fedot.readthedocs.io/en/master/advanced/hyperparameters_tuning.html>_ are supported including models' custom evaluation metrics and search spaces;
  • Versatility. FEDOT is not limited to specific modeling tasks <https://fedot.readthedocs.io/en/master/advanced/architecture.html>_, for example, it can be used in ODE or PDE;
  • Reproducibility. Resulting pipelines can be exported separately as JSON <https://fedot.readthedocs.io/en/master/advanced/pipeline_import_export.html>_ or together with your input data as ZIP archive <https://fedot.readthedocs.io/en/master/advanced/project_import_export.html>_ for experiments reproducibility;
  • Customizability. FEDOT allows managing models complexity <https://fedot.readthedocs.io/en/master/introduction/fedot_features/automation_features.html#models-used>_ and thereby achieving desired quality.

Installation

  • Package installer for Python pip

The simplest way to install FEDOT is using pip:

.. code-block::

$ pip install fedot

Installation with optional dependencies for image and text processing, and for DNNs:

.. code-block::

$ pip install fedot[extra]

  • Docker container

Available docker images can be found here here <https://github.com/aimclub/FEDOT/tree/master/docker/README_en.rst>_.

How to Use

FEDOT provides a high-level API that allows you to use its capabilities in a simple way. The API can be used for classification, regression, and time series forecasting problems.

To use the API, follow these steps:

  1. Import Fedot class

.. code-block:: python

from fedot.api.main import Fedot

  1. Initialize the Fedot object and define the type of modeling problem. It provides a fit/predict interface:
  • Fedot.fit() begins the optimization and returns the resulting composite pipeline;
  • Fedot.predict() predicts target values for the given input data using an already fitted pipeline;
  • Fedot.get_metrics() estimates the quality of predictions using selected metrics.

NumPy arrays, Pandas DataFrames, and the file's path can be used as sources of input data. In the case below, x_train, y_train and x_test are numpy.ndarray():

.. code-block:: python

model = Fedot(problem='classification', timeout=5, preset='best_quality', n_jobs=-1)
model.fit(features=x_train, target=y_train)
prediction = model.predict(features=x_test)
metrics = model.get_metrics(target=y_test)

More information about the API is available in the documentation <https://fedot.readthedocs.io/en/latest/api/api.html>__ section and advanced approaches are in the advanced <https://github.com/aimclub/FEDOT/tree/master/examples/advanced>__ section.

Examples & Tutorials

Jupyter notebooks with tutorials are located in examples repository <https://github.com/ITMO-NSS-team/fedot-examples>__. There you can find the following guides:

  • Intro to AutoML <https://github.com/ITMO-NSS-team/fedot-examples/blob/main/notebooks/latest/1_intro_to_automl.ipynb>__
  • Intro to FEDOT functionality <https://github.com/ITMO-NSS-team/fedot-examples/blob/main/notebooks/latest/2_intro_to_fedot.ipynb>__
  • Intro to time series forecasting with FEDOT <https://github.com/ITMO-NSS-team/fedot-examples/blob/main/notebooks/latest/3_intro_ts_forecasting.ipynb>__
  • Advanced time series forecasting <https://github.com/ITMO-NSS-team/fedot-examples/blob/main/notebooks/latest/4_auto_ts_forecasting.ipynb>__
  • Gap-filling in time series and out-of-sample forecasting <https://github.com/ITMO-NSS-team/fedot-examples/blob/main/notebooks/latest/5_ts_specific_cases.ipynb>__
  • Hybrid modelling with custom models <https://github.com/ITMO-NSS-team/fedot-examples/blob/main/notebooks/latest/6_hybrid_modelling.ipynb>__

Notebooks are issued with the corresponding release versions (the default version is 'latest').

Also, external examples are available:

  • Kaggle: baseline for Microsoft Stock - Time Series Analysis task <https://www.kaggle.com/dreamlone/microsoft-stocks-price-prediction-automl>__

Extended examples:

  • Credit scoring problem, i.e. binary classification task <https://github.com/aimclub/FEDOT/blob/master/examples/real_cases/credit_scoring/credit_scoring_problem.py>__
  • Time series forecasting, i.e. random process regression <https://github.com/aimclub/FEDOT/blob/master/examples/real_cases/metocean_forecasting_problem.py>__
  • Spam detection, i.e. natural language preprocessing <https://github.com/aimclub/FEDOT/blob/master/examples/real_cases/spam_detection.py>__
  • Wine variety prediction with multi-modal data <https://github.com/aimclub/FEDOT/blob/master/examples/advanced/multimodal_text_num_example.py>__

Also, several video tutorials are available available <https://www.youtube.com/playlist?list=PLlbcHj5ytaFUjAxpZf7FbEaanmqpDYhnc>__ (in Russian).

Publications About FEDOT

We also published several posts devoted to different aspects of the framework:

In English:

  • How AutoML helps to create composite AI? - towardsdatascience.com <https://towardsdatascience.com/how-automl-helps-to-create-composite-ai-f09e05287563>__
  • AutoML for time series: definitely a good idea - towardsdatascience.com <https://towardsdatascience.com/automl-for-time-series-definitely-a-good-idea-c51d39b2b3f>__
  • AutoML for time series: advanced approaches with FEDOT framework - towardsdatascience.com <https://towardsdatascience.com/automl-for-time-series-advanced-approaches-with-fedot-framework-4f9d8ea3382c>__
  • Winning a flood-forecasting hackathon with hydrology and AutoML - towardsdatascience.com <https://towardsdatascience.com/winning-a-flood-forecasting-hackathon-with-hydrology-and-automl-156a8a7a4ede>__
  • Clean AutoML for “Dirty” Data - towardsdatascience.com <https://towardsdatascience.com/clean-automl-for-dirty-data-how-and-why-to-automate-preprocessing-of-tables-in-machine-learning-d79ac87780d3>__
  • FEDOT as a factory of human-competitive results - youtube.com <https://www.youtube.com/watch?v=9Rhqcsrolb8&ab_channel=NSS-Lab>__
  • Hyperparameters Tuning for Machine Learning Model Ensembles - towardsdatascience.com <https://towardsdatascience.com/hyperparameters-tuning-for-machine-learning-model-ensembles-8051782b538b>__

In Russian:

  • Как AutoML помогает создавать модели композитного ИИ — говорим о структурном обучении и фреймворке FEDOT - habr.com <https://habr.com/ru/company/spbifmo/blog/558450>__
  • Прогнозирование временных рядов с помощью AutoML - habr.com <https://habr.com/ru/post/559796/>__
  • Как мы “повернули реки вспять” на Emergency DataHack 2021, объединив гидрологию и AutoML - habr.com <https://habr.com/ru/post/577886/>__
  • Чистый AutoML для “грязных” данных: как и зачем автоматизировать предобработку таблиц в машинном обучении - ODS blog <https://habr.com/ru/company/ods/blog/657525/>__
  • Фреймворк автоматического машинного обучения FEDOT (Конференция Highload++ 2022) - presentation <https://docs.yandex.ru/docs/view?url=ya-disk-public%3A%2F%2Fi27LScu3s3IIHDzIXt9O5EiEAMl6ThY6QLu3X1oYH%2FFiAl%2BLcNp4O4yTSYd2gRZnW5aDQ4kMZEXE%2BwNjbq78ug%3D%3D%3A%2F%D0%94%D0%B5%D0%BD%D1%8C%201%2F4.%D0%A1%D0%B8%D0%BD%D0%BD%D0%B0%D0%BA%D1%81%2F9.Open%20source-%D1%82%D1%80%D0%B8%D0%B1%D1%83%D0%BD%D0%B0_HL_FEDOT.pptx&name=9.Open%20source-%D1%82%D1%80%D0%B8%D0%B1%D1%83%D0%BD%D0%B0_HL_FEDOT.pptx>__
  • Про настройку гиперпараметров ансамблей моделей машинного обучения - habr.com <https://habr.com/ru/post/672486/>__

In Chinese:

  • 生成式自动机器学习系统 (presentation at the "Open Innovations 2.0" conference) - youtube.com <https://www.youtube.com/watch?v=PEET0EbCSCY>__

Project Structure

The latest stable release of FEDOT is in the master branch <https://github.com/aimclub/FEDOT/tree/master>__.

The repository includes the following directories:

  • Package core <https://github.com/aimclub/FEDOT/tree/master/fedot/core>__ contains the main classes and scripts. It is the core of the FEDOT framework
  • Package examples <https://github.com/aimclub/FEDOT/tree/master/examples>__ includes several how-to-use-cases where you can start to discover how FEDOT works
  • All unit and integration tests can be observed in the test <https://github.com/aimclub/FEDOT/tree/master/test>__ directory
  • The sources of the documentation are in the docs <https://github.com/aimclub/FEDOT/tree/master/docs>__ directory

Current R&D and future plans

Currently, we are working on new features and trying to improve the performance and the user experience of FEDOT. The major ongoing tasks and plans:

  • Implementation of large language model for AutoML tasks in FEDOT.LLM <https://github.com/ITMO-NSS-team/FEDOT.LLM>__.
  • Implementation of meta-learning based at GNN and RL (see GAMLET <https://github.com/ITMO-NSS-team/GAMLET>__).
  • Improvement of the optimisation-related algorithms implemented in GOLEM <https://github.com/aimclub/GOLEM/>__.
  • Support for more complicated pipeline design patters, especially for time series forecasting.

In addition, we are working on a number of research tasks related to benchmarking of time series forecasting using AutoML and multimodal modelling.

Any contribution is welcome. Our R&D team is open for cooperation with other scientific teams as well as with industrial partners.

Documentation

Also, a detailed FEDOT API description is available in Read the Docs <https://fedot.readthedocs.io/en/latest/>__.

Contribution Guide

  • The contribution guide is available in this repository <https://github.com/aimclub/FEDOT/blob/master/docs/source/contribution.rst>__.

Acknowledgments

We acknowledge the contributors for their important impact and the participants of numerous scientific conferences and workshops for their valuable advice and suggestions.

Funding

This research is financially supported by the Foundation for National Technology Initiative's Projects Support as a part of the roadmap implementation for the development of the high-tech field of Artificial Intelligence for the period up to 2030 (agreement 70-2021-00187)

Side Projects

  • The optimisation core implemented in the GOLEM.
  • The prototype of the web-GUI for FEDOT is available in the FEDOT.WEB <https://github.com/aimclub/FEDOT.Web>__ repository.
  • The prototype of FEDOT-based meta-AutoML in the GAMLET.
  • The prototype of FEDOT-based LLM for nexgen AutoML in the FEDOT.LLM.

Contacts

  • Telegram channel for solving problems and answering questions about FEDOT <https://t.me/FEDOT_helpdesk>_
  • Natural System Simulation Team <https://itmo-nss-team.github.io/>_
  • Anna Kalyuzhnaya <https://scholar.google.com/citations?user=bjiILqcAAAAJ&hl=ru>_, Team leader (anna.kalyuzhnaya@itmo.ru)
  • Telegram newsfeed channel <https://t.me/NSS_group>_
  • Youtube channel <https://www.youtube.com/channel/UC4K9QWaEUpT_p3R4FeDp5jA>_

Supported by

  • National Center for Cognitive Research of ITMO University <https://actcognitive.org/>_

Citation

@article{nikitin2021automated, title = {Automated evolutionary approach for the design of composite machine learning pipelines}, author = {Nikolay O. Nikitin and Pavel Vychuzhanin and Mikhail Sarafanov and Iana S. Polonskaia and Ilia Revin and Irina V. Barabanova and Gleb Maximov and Anna V. Kalyuzhnaya and Alexander Boukhanovsky}, journal = {Future Generation Computer Systems}, year = {2021}, issn = {0167-739X}, doi = {https://doi.org/10.1016/j.future.2021.08.022}}

@inproceedings{polonskaia2021multi, title={Multi-Objective Evolutionary Design of Composite Data-Driven Models}, author={Polonskaia, Iana S. and Nikitin, Nikolay O. and Revin, Ilia and Vychuzhanin, Pavel and Kalyuzhnaya, Anna V.}, booktitle={2021 IEEE Congress on Evolutionary Computation (CEC)}, year={2021}, pages={926-933}, doi={10.1109/CEC45853.2021.9504773}}

.. |docs| image:: https://readthedocs.org/projects/ebonite/badge/?style=flat :target: https://fedot.readthedocs.io/en/latest/ :alt: Documentation Status

.. |build| image:: https://github.com/aimclub/FEDOT/actions/workflows/unit-build.yml/badge.svg :alt: Build Status :target: https://github.com/aimclub/FEDOT/actions/workflows/unit-build.yml

.. |integration| image:: https://github.com/aimclub/FEDOT/actions/workflows/integration-build.yml/badge.svg :alt: Integration Build Status :target: https://github.com/aimclub/FEDOT/actions/workflows/integration-build.yml

.. |coverage| image:: https://codecov.io/gh/aimclub/FEDOT/branch/master/graph/badge.svg :alt: Coverage Status :target: https://codecov.io/gh/aimclub/FEDOT

.. |pypi| image:: https://badge.fury.io/py/fedot.svg :alt: Supported Python Versions :target: https://badge.fury.io/py/fedot

.. |python| image:: https://img.shields.io/pypi/pyversions/fedot.svg :alt: Supported Python Versions :target: https://img.shields.io/pypi/pyversions/fedot

.. |license| image:: https://img.shields.io/github/license/aimclub/FEDOT :alt: Supported Python Versions :target: https://github.com/aimclub/FEDOT/blob/master/LICENSE.md

.. |downloads_stats| image:: https://static.pepy.tech/personalized-badge/fedot?period=total&units=international_system&left_color=grey&right_color=brightgreen&left_text=Downloads :target: https://pepy.tech/project/fedot

.. |tg| image:: https://img.shields.io/badge/Telegram-Group-blue.svg :target: https://t.me/FEDOT_helpdesk :alt: Telegram Chat

.. |ITMO| image:: https://raw.githubusercontent.com/aimclub/open-source-ops/43bb283758b43d75ec1df0a6bb4ae3eb20066323/badges/ITMO_badge_rus.svg :alt: Acknowledgement to ITMO :target: https://itmo.ru

.. |NCCR| image:: https://raw.githubusercontent.com/aimclub/open-source-ops/43bb283758b43d75ec1df0a6bb4ae3eb20066323/badges/NCCR_badge.svg :alt: Acknowledgement to NCCR :target: https://actcognitive.org/

.. |gitlab| image:: https://img.shields.io/badge/mirror-GitLab-orange :alt: GitLab mirror for this repository :target: https://gitlab.actcognitive.org/itmo-nss-team/FEDOT