Groups
Category
Level
Minimum Description Length (MDL) picks the model that compresses the data best by minimizing L(M) + L(D|M).
The bias–variance tradeoff explains how prediction error splits into bias squared, variance, and irreducible noise for squared loss.