Auto-sklearn is an AutoML framework that is based on scikit-Learn. This open-source platform combines powerful techniques and methods to help overcome a wide range of AutoML challenges. 

It mainly solves regression and classification problems. The first version was introduced in the article “Efficient and robust automated machine learning” in 2015. The second version launched in 2020. The product’s three main modules are meta-learning, build ensemble, and Bayesian optimization.

Auto-sklearn extends the configuration of the general ML framework with global optimization introduced with Auto-WEKA. It helps enhance generalization by creating an ensemble of all models tested at the time of global optimization process.

At present, Auto-sklearn consists of total of 15 classification algorithms and 14 feature pre-processing algorithms. It also handles data scaling and encoding of categorical parameters and missing values. 

Project Background

• • Library: Auto-sklearn AutoML library 
  • Author: Matthias Feurer
  • Initial Release: 2016
  • Type:  Out of the box supervised machine learning
  • License: 3-clause BSD license
  • Contains: Modules, metrics, extension interfaces
  • Language: Python
  • GitHub: /automl/auto-sklearn
  • Runs On: Linux
  • Stackflow: /questions/tagged/auto-sklearn

Applications

• • Preprocessing
  • Model selection
  • Hyperparameter tuning
  • Classification 
  • Regression