Stable Forward Search for Feature Selection
Défense de mémoire de Michel Gauthier
Date : 28/08/2018 15:00 - 28/08/2018 16:00
Lieu : salle académique
Orateur(s) : Michel Gauthier
Organisateur(s) : Isabelle Daelman
In order use machine learning, a model needs to be trained based on a dataset characterized by a feature set. This set can contain numerous and not always useful features for the model. The feature
selection can help to sort features and reduce the set of features. The goal is to take the most useful of them and maximize the learning phase of the model. There are diff erent kinds of selections based on diff erent
methods that off er various benefi ts like rigorous or execution speed. The number of features selected need to be smaller to reduce the computation
time and the complexity without losing information. The utilization of a stopping criteria is important and can make the di fference between a relevant or irrelevant selection.
Contact :
Isabelle Daelman
-
4966
-
isabelle.daelman@unamur.be
Télecharger :
vCal