Data, Ensemble Prediction & Residuals
Data Points
Ensemble Prediction
True Function
Current Tree Fitting Residuals
Residuals to Fit
Tree Prediction
Training Loss Over Iterations
η = 0.3
depth = 2
n = 50
Gradient Boosting
F₀(x) = ȳ (initial prediction)
For m = 1 to M:
r_im = y_i - F_{m-1}(x_i)
fit tree h_m to residuals
F_m = F_{m-1} + η · h_m
For m = 1 to M:
r_im = y_i - F_{m-1}(x_i)
fit tree h_m to residuals
F_m = F_{m-1} + η · h_m
Current State
Trees:
0
MSE:
-
R²:
-
Viewing Tree:
-
Current Tree Splits
No trees yet