Skip to content

GBDT(Gradient Boosting Decision Tree)

Widely-used machine learning algorithm, due to its efficiency, accuracy, and interpretability. It is facing chanllenges, especially in the tradeoff between accuracy and efficiency. GBDT is a machine learning technique that combines two core ideas: - Gradient Boosting - Decision Tree

Process

  1. Starting with a basic guess
  2. Building trees sequentially, where each new tree corrects the errors of the previous ones
  3. Using gradients to guide the process

Key Features

  • Weak Learners
  • Sequential
  • Learning Rate
  • Additive

GBDT vs Random Forest

RF builds trees independently and average them, GBDT builds them sequentially to fix errors. GBDT doesn't assume linearity or specific distributions.

Difficult

It's unclear how to perform data sampling