The expected squared-error of prediction
E(Y,ˆθ_)|x∗_{Y−ˆY}2=EY|x_∗{Y−f(θ;x_∗)}2⏟Variability of Y around its conditional expected value+[f(θ;x_∗)−Eˆθ|x_∗{f(ˆθ_;x_∗)}]2⏟Difference between the expected value and its estimate, Squared Bias+Eˆθ_|x_∗[f(ˆθ_;x_∗)−Eˆθ_|x_∗{f(ˆθ_;x_∗)}]2⏟Change in the model due to the training data used, Variance