Differences

This shows you the differences between two versions of the page.

--- data_mining:gradient_boosting [2016/02/01 23:04] – phreazer
+++ data_mining:gradient_boosting [2017/07/19 20:34] (current) – [Basic idea] phreazer
@@ Line 42: / Line 42: @@
 Beispielhafte Loss-Funktion: Quadrierter Fehler in Gauß'scher Regression $p=(Y-f(X))^2$
+Gegeben $X=(X_1, ..., X_n), Y=(Y_1,...,Y_n)$, minimiere das empirische Risiko
+$$R = \frac{1}{n} \sum_{i=1}^n p(Y_i, f(X_i))$$
+Beispiel (quadratischer Fehler):
+$$R = \frac{1}{n} \sum_{i=1}^n p(Y_i - f(X_i))^2$$
+===== Basic idea =====
+  - Generate simple regression model.
+  - Compute error residual.
+  - Learn to predict residual
+  - Add to initial model