When constructing random forest, one way is to take the simple average of all trees' predictions. Alternatively, we can also calculate the weight assigned to each tree by a function of error rate. However, is it a kind of boosting?
I originally think boosting must be sequentially updating the focus to wrongly classified samples during training but I am not sure is it a must. Thanks!
Boosted Random Forest(p3): http://www.vision.cs.chubu.ac.jp/MPRG/C_group/C058_mishina2014.pdf