1: \begin{definition}[{\bf Main Theorem}]

2: \label{Theorem1} Consider the previous protocol for randomness

3: amplification and the conditional probability distribution $P (k,

4: \tilde y,t,g,e|z)$ describing the statistics of the bits $k,

5: \tilde y, t, g$ generated during its execution and any possible

6: system with input $z$ and output $e$ correlated to them. The

7: probability $P({\rm guess})$ of correctly guessing between this

8: distribution and the ideal distribution $P_{\rm ideal} (k, \tilde

9: y, t, g, e|z)$ is such that

10: \begin{equation}

11: \label{t1}

12:     P({\rm guess})

13: \ \leq\

14:     \frac 1 2 + \frac{3 \sqrt{N_d}}{2}

15:     \left[ \alpha^{N_d} +

16:     2\, N_b^{\log_2(1-\epsilon)}

17:     \left( 32\beta \epsilon^{-5} \right)^{N_d} \right]\ .

18: \end{equation}

19: where $\alpha$ and $\beta$ are real numbers such that % defined in Lemma~\ref{num}

20: $0< \alpha <1 <\beta$.

21: \end{definition}

22: