Interpretable AI MEAP V02 by Ajai Thampi

Author:Ajai Thampi [Thampi, Ajai]
Language: eng
Format: epub, mobi
Publisher: Manning Publications Co.
Published: 2020-07-27T22:00:00+00:00

Figure 3.6: Illustration of the Random Forest Algorithm

For completeness, let’s also learn how the adaptive boosting and gradient boosting algorithms work. Adaptive boosting, usually shortened as AdaBoost, is illustrated in Figure 3.7. The algorithm works as follows. The first step is to train a decision tree using all the training data. Each data point is given equal weight for the first decision tree. Once the first decision tree is trained, calculate the error rate of the tree by taking a weighted sum of the error for each data point. This weighted error rate is then used to determine the weight of the decision tree. If the error rate of the tree is high, then a lower weight is given for the tree since its predictive power is low. If the error rate is low, then a higher weight is given for the tree since it has a higher predictive power. The weight of the first decision tree is then used to determine the weights for each data point for the second decision tree. The wrongly classified data points will be given a higher weight so that the second decision tree can try to reduce the error rate. This process is then repeated in sequence until the number of trees we set during training is reached. After all the trees have been trained, then we come up with the final prediction by taking a weighted majority vote. Since a decision tree with a higher weight has higher predictive power, it is given more influence in the final prediction.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.