In the previous chapter, we saw the power of multiple weak learners that can do magic and learn nonlinear data. We discussed boosting and saw how it can be used to solve extremely complex problems such as face detection, and we it did quite well. I just want to repeat the points we followed in the AdaBoost algorithms before moving ahead:
- Loading data and weight each instance equally
- Training a weak learner (we used decision stump) on the data
- Evaluating the errors made by the classifier and giving more weight to wrongly classified instances
- Repeating the procedure from point to for number of iterations
- Using the error rate of the weak learner as the weight of prediction made by the classifier when learning is complete ...