Tree-Based Pipeline Optimization

The full form TPOT is Tree-Based Pipeline Optimization. It’s one of the very first and most useful open-source software packages in the community. Dr. Randal Olson developed TPOT and is still working on it to expand is functionality.

The primary goal of TPOT is to automate the process of building ML pipelines. For this purpose, you have to combine a flexible expression tree pipelines representations with stochastic search algorithms. The platform uses Python-based scikit-learn library that serves as its ML menu.

TPOT takes a while to execute on larger datasets. It’s because of TPOT’s default settings that make it evaluate 10,000 pipeline configurations prior to finishing.

Some of its features include TPOT with code, TPOT on the command line, template options, pipeline coaching, built-in configurations, FeatureSetSelector, and many more.

Project Background

    • Library: TPOT AutoML tool
    • Author: Dario Radečić
    • Initial Release: 2020
    • Type: Initially developed for the science community
    • License: LGPL-3.0 License
    • Contains: Scikit-learn data preparation and machine learning models
    • Language: Python, Jupyter Notebook, Shell
    • GitHub: /EpistasisLab/tpot
    • Runs On: Linux, Mac, Windows
    • GitHub Discussions: None
    • Stackflow: /questions/tagged/tpot

Applications

  • Classification
  • Digits dataset
  • Regression
Scroll to Top