Mathematics Winter School 

14/03/2024, 11:30 — 11:45 — Room P3.10, Mathematics Building Online
Rosário Oliveira, Instituto Superior Técnico

What is relevant and what is redundant?

Feature selection is recognized as an important preprocessing technique to reduce dimensionality and improve the performance of regression and classification tasks. The class of sequential forward feature selection methods based on Mutual Information is widely used, mainly due to its computational efficiency and independence from the specific classifier. In this talk, we present a theoretical framework for this class of methods which explains the existing proposals as approximations of an optimal target objective function, that quantifies redundancy, relevance, and irrelevance of the features for the classification problem.

See also

11H30_MROS.pdf