LOOCV
Leave-one-out cross-validation is the extreme form of K-fold cross-validation in which K equals the sample size N. Each iteration holds out a single observation as the test set while the remaining observations are used for training. The procedure is repeated N times and the average error is taken. For linear models, a closed-form solution exists.
Timeline (last 90 days)
The paper's analytical framework is based on leave-one-out cross-validation (LOOCV), and derives the expected out-of-sample performance for OLS and ridge regression separately.
In OLS, the LOOCV residual can be written as e_i/(1-h_ii), where h_ii is exactly the leverage value, which explains why D* naturally enters the formula for expected out-of-sample performance.
LOOCV has high variance on high-leverage data, and may overestimate generalization performance when observations are correlated.
All Facts (3)
The paper's analytical framework is based on leave-one-out cross-validation (LOOCV), and derives the expected out-of-sample performance for OLS and ridge regression separately.
50%UnverifiedIn OLS, the LOOCV residual can be written as e_i/(1-h_ii), where h_ii is exactly the leverage value, which explains why D* naturally enters the formula for expected out-of-sample performance.
50%UnverifiedLOOCV has high variance on high-leverage data, and may overestimate generalization performance when observations are correlated.
50%