Question d'entretien d'embauche BNP Paribas: Bagging ? Boosting ?... | Glassdoor.fr

Question d'entretien d'embauche

Entretien de Data Scientist Paris

Bagging ? Boosting ?

Répondre

Réponse de l'entretien

1 réponse

0

In bagging technique, data set is divided into subsets using sampling with replacement, then a model is trained on each subset. The final prediction is given by voting or averaging result of the models. Bagging is done in parallel. Boosting, however, is sequential. After each round of prediction, the mis-predicted results get higher weight, so that they gain more attention in the next round of training.
Both bagging and boosting are ensemble learning strategies.

Wang Siyun, le 11 mai 2018

Ajouter des réponses ou des commentaires

Pour commenter ceci, se connecter ou s'inscrire