Computer scienceData scienceMachine learningEnsemble learning

Gradient boosting

Minimum number of samples

Report a typo

Suppose we trained gradient boosting with the learning rate of 0.1 for 500 rounds and the optimal tree depth in the regression setting. We want to determine which minimum number of samples in the leaf will help to improve the performance:

A box plot of 5, 9 and 11 minimum sample in the leafes, with the minimum number of samples on the X axis and MSE on the Y axis

Based on the plot above, what is the optimal value for the minimum samples in the leaf?

Select one option from the list

Create a free account to access the full topic