A Theorem Proofs

In this section, we present the proofs to the theorems introduced in the main paper. The proof to Theorem 2 is presented as follows. Consider a classification task where the loss function is the cross entropy loss. This approximately holds for many applications with over-parameterized neural predictors. In this case, we have the following theorem: Theorem 3. If Equations (18) and (19) hold, that This contradicts with Equation (23).