I have read that in ensemble learning we use the outputs of various classifiers to make the predictive modeling better but in Adaboost we just use one classifier and we make it a strong learner but how it is a part of ensemble learning.
AdaBoost or Adaptive Boost is a boosting ensemble model which works by learning from it's previous mistakes, ie: misclassified data points.
We specify the number of decision trees to be generated while training and during each training step, it calculates the following :
- The weighted error rate of the trained decision tree
- The decision tree's weight in the ensemble = learning rate * log((1-e)/e)
- Update weights of wrongly classified points
This process repeats until all the trees are trained. In the end, AdaBoost makes all of them "vote" by adding up the weight (of each tree) multiplying by the prediction (of each tree).
Therefore, this becomes an ensemble of multiple decision trees.