To comprehend the generalization power of ensemble learning being superior to an individual learner, Dietterich provided three reasons.
These three reasons help us understand the reason for the superiority of ensemble learning leading to a better hypothesis:
The training information won't give adequate data to picking a single best learner. For instance, there might be numerous learners performing similarly well on the training information set. In this way, joining these learners might be a superior decision.
The second reason is that, the search procedures of the learning algorithms may be defective. For instance, regardless of the possibility that there exists a best hypothesis, the learning algorithms may not be able to achieve that due to various reasons including generation of an above average hypothesis. Ensemble learning can improve on that part by increasing the possibility to achieve the best hypothesis.
The third reason is that one target function...