Thuật ngữ AI
Từ điển đầy đủ về Trí tuệ nhân tạo
Functional Gradient Descent
Optimization technique that adjusts model parameters in the direction opposite to the gradient of the loss function in function space.
Gradient Boosting Machines (GBM)
Class of algorithms implementing the gradient boosting method with various loss functions and weak learners.
Bagging vs Boosting
Fundamental distinction where bagging trains models independently in parallel while boosting builds them sequentially by correcting errors.
Loss function gradient
Partial derivative of the loss function with respect to predictions, indicating the direction of greatest possible improvement.
Regularization in GBM
Set of techniques (L1/L2, tree constraints, shrinkage) controlling model complexity to prevent overfitting.
Newton's method in GBM
Variant of gradient boosting using the Hessian matrix (second derivative) for more optimal parameter updates.