Deep Learning Ensembling

Question 1

Bagging, Boosting, Stacking

Answer 1

Bagging (Bootstrap Aggregation): Multiple subsets (sampling with replacement) -> multiple classifiers –> majority votes. Decreases variance, e.g., random forest

Boosting: Sequential, after each stage adjust the weight of samples

Question 2

Bootstrapping

Answer 2

Sample multiple sub-sets with replacement