Overview of integrated learning

Integrated learning

Bagging(Bootstrap aggregating)

The main idea ：

1. Want to get the integration with strong generalization , Each weak classifier should be as independent as possible
2. Each weak classifier samples different training sets , Make weak classifiers as different as possible
   

Stacking

The main idea ：

1. The predicted value will be used as the characteristic value of the training sample , Finally, we will get new training samples , Then the model is obtained by training based on the new training samples , Then get the final prediction result .
   

Boosting

The main idea ：

1. The sample that the last classifier classified incorrectly , Follow up focus on strengthening learning .
   

2. Adboost Detailed explanation

Generative and discriminant

1. Generative model ： Learn the internal distribution of each category , Calculate the probability of belonging to this category according to the distribution .
2. Discriminant model ： Learn the boundaries of different categories