Gradient Boosting Decision Tree (GBDT)及其实现细节与优化版本分析

梯度提升决策树

The term for Gradient Boosting Decision Tree (GBDT) in English is an iterative decision tree algorithm composed of multiple decision trees. The algorithm aggregates the results from all trees to produce the final output. It is important to note that GBDT consists of regression trees rather than classification trees. This method is predominantly used for regression tasks, including both linear and nonlinear problems, although it can also be applied to classification tasks.

我们已经在弄明白决策树(DT)部分之后

输入

训练集数据D由有序对(x_i, y_i)构成（i从1到N）。为了衡量预测值f(x)与真实值y之间的差异程度，在此研究中我们采用了标准损失函数L(y, f(x