Open source AutoML: machine learning for everyone?

Author

•  Andrea Lanari

Abstract

During the past decade, Machine Learning (ML) popularity has been skyrocketing: year by year, the models are becoming more and more ac- curate and they are finding applications in a huge variety of fields. Nowa- days, even if ML is widely used, developing a good Machine Learning system is not so easy: in particular, the model choice and hyperparam- eters tuning are two aspects that only experts can handle successfully through many tests. Therefore, researchers in the field are tackling this problem making use of a new method: Automated Machine Learning (Au- toML). AutoML is the possibility to delete the “expert phase” in favor of a more complex system able to find the best approach independently. In this work, we focus on the open source tools already available, trying to compare their performance and their actual autonomy.

Outcomes

• project forum
• repository
• final report
• slides