I will present a finite-sample generalization analysis of two-part code MDL estimator. This method selects a model that minimizes the sum of the model description length plus the data description leng
