Automated machine learning (AutoML) is a hot new field with the goal of making it easy to select machine learning algorithms, their parameter settings, and the pre-processing methods that improve their ability to detect complex patterns in big data. The field got started in earnest around 2015 and is made possible by mature high-performance computing technologies and a number of open-source libraries for ML including the Python-based scikit-learn resource that includes a comprehensive suite of ML algorithms and methods for data pre-processing.
We have developed two AutoML methods with open-source software packages. Our first method is the Tree-Based Pipeline Optimization (TPOT) method that uses binary expression trees to represent ML pipelines with optimization provided by genetic programming and other stochastic search methods. TPOT resources are outlined here.
We are also developing the University of Pennsylvania accessible Artificial Intelligence (PennAI) system that learns from experience how to select ML methods and parameter settings for data analysis. The PennAI method and software is also summarized here.
There are several other AutoML methods and software packages that have been developed. For example, Auto-Weka uses as its base the popular Weka package for ML. Another example is the Robust Bayesian Optimization (RoBO) framework. Here is Auto-Sklearn. Here is AutoKeras for deep learning. Google is now providing AutoML services as well. There is a new book that reviews a number of AutoML methods including TPOT.
Please contact us if you would like your AutoML algorithm and software listed here.