I broadly understand Boosting as focusing and putting more weight on weak learners and combining multiple weak learners to have a more accurate prediction
I understand in case of decision trees , how boosting works by constructing multiple decision trees ( an ensemble ) , focusing on the weak decision stumps and then predicting outcomes by optimizing on the weights on the learners
But is boosting an add on or a method in itself ? I hear it being applied to any modelling technique ? like logistic regression. How does it work there ? What is the intuitive or simple analogy to understand boosting over an existing modelling algorithm , even linear regression ?