Shrinkage
الانكماش
تقنية تضرب إسهام كل شجرة جديدة بعامل صغير (معدل التعلم η) قبل إضافتها للتجميعة، مما يترك مجالاً للأشجار اللاحقة ويقلّل فرط التخصيص.
A technique that scales each new tree's contribution by a small factor (learning rate η) before adding it to the ensemble, leaving room for future trees and reducing overfitting.
Also translated asمعامل الانكماش، تقليص الإسهام، إبطاء التعلّم
First appears in this corpus in: Ridge Regression: Biased Estimation for Nonorthogonal Problems (1970)
Appears in these papers
- A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting1997in the sky ✦
- Greedy Function Approximation: A Gradient Boosting Machine2001in the sky ✦
- Greedy Function Approximation: A Gradient Boosting Machine2001in the sky ✦
- Regression Shrinkage and Selection via the Lasso1996in the sky ✦
- Ridge Regression: Biased Estimation for Nonorthogonal Problems1970in the sky ✦
- XGBoost: A Scalable Tree Boosting System2016in the sky ✦