MFE: Towards reproducible meta-feature extraction

Edesio Alcobaça, Felipe Siqueira, Adriano Rivolli, Luis Paulo Faina Garcia, Jefferson T. Oliva, Andre Carlos Ponce de Leon Ferreira De Carvalho

Link To Paper: https://www.jmlr.org/papers/volume21/19-348/19-348.pdf

Journal Of Paper: Journal of Machine Learning Research (JMLR)

Confirmed Open Access: Yes

Topics From Call For Papers:

The paper falls under the following topics:

* Meta-Learning and Learning to learn
* Reproducibility

Broader Impact Statement On Ethical And Societal Implications:

In this paper, researchers proposed an alternative to deal with the difficulty of using general frameworks for meta-feature extraction and reproducibility by developing a meta-feature extractor package. The paper introduces tools following uniform guidelines that facilitate the use and inclusion of new meta-features. Two packages are presented, one developed in python called pyMFE and another in R called MFE.

As the paper presented tools, there are no high-risk ethical and societal implications. However, AutoML tools could use our proposed tools, and consequently, they could add some bias that we cannot measure.