1: \begin{proof}

2: The lemma essentially follows from the symmetry/exchangeability property

3: of Models~\ref{mdl:Z2} (Z2) and \ref{mdl:CBM} (CBM). Rigorous proof

4: of this intuitive result is however quite technical, as the definition

5: of clustering error involves a global sign flipping. Fortunately,

6: most of the work has been done in~\cite{zhang2016minimax}. In particular,

7: note that the parameter space $\paramset$ is closed under permutation

8: in the sense that for any label vector $\Label\in\paramset$ and any

9: permutation $\pi$ on $\left[\num\right]$, the new label vector $\Label'$

10: defined by $\labels'(i)\coloneqq\labels(\pi^{-1}(i))$ also belongs

11: to $\paramset$. It can also be seen that both Models~\ref{mdl:Z2}

12: (Z2) and~\ref{mdl:CBM} (CBM) are homogeneous, i.e., the distribution

13: of each $\adj_{ij}$ is uniquely determined by the sign of $\labelstar_{i}\labelstar_{j}$.

14: Consequently, for Model \ref{mdl:CBM} (CBM) this lemma immediately

15: follows from Lemma 2.1 in \cite{zhang2016minimax}, as its proof applies

16: without change. For Model \ref{mdl:Z2} (Z2) in which the distribution

17: of $\Adj$ is continuous, we note that the proof of Lemma 2.1 in \cite{zhang2016minimax}

18: continues to hold when summations therein are replaced by appropriate

19: integrations. 

20: \end{proof}