1: \begin{definition}

2: 	\label{def:same}

3: 	Let $[\bs G_1, \bs o_1]$ and $[\bs G_2, \bs o_2]$ be (not necessarily unimodular) random rooted networks. The following conditions are different definitions for $[\bs G_1,\bs o_1]$ and $[\bs G_2,\bs o_2]$ to be \defstyle{unrooted-equivalent} (or to \defstyle{have the same non-rooted networks}).

4: 	

5: 	\begin{description}

6: 		\item [(B)\label{conditionB}]

7: 		%\mar{"each one" is important}

8: 		The distribution of each one is obtained from the other by a \textit{biasing} and then a \textit{root-change}.

9: 		\item [(R)\label{conditionR}] 

10: 		The distribution of $[\bs G_2,\bs o_2]$ is obtained from $[\bs G_1,\bs o_1]$ by a {root-change}.

11: 		\item [(C)\label{conditionC}]

12: 		%\mar{doubt: isn't there a measurability issue for \textit{selecting} the isomorphism? Answer: No!}

13: 		There is a coupling of them (i.e. a probability measure on $\mathcal G_*\times\mathcal G_*$ whose marginals are identical with the distributions of $[\bs G_i,\bs o_i]$'s) which is concentrated on the set of pairs of rooted networks with the same non-rooted networks; i.e. $\{([G_1,o_1],[G_2,o_2]): [G_1]=[G_2]\}$. 

14: 		\item [(D)\label{conditionD}]

15: 		There is a random doubly-rooted  network $[\bs G, \bs o, \bs o']$ such that $[\bs G, \bs o]$ and $[\bs G, \bs o']$ have the same distributions as $[\bs G_1, \bs o_1]$ and $[\bs G_2, \bs o_2]$ respectively.

16: 		\item [(F)\label{conditionF}]

17: 		By forgetting the roots, the random non-rooted networks $[\bs G_1]$ and $[\bs G_2]$ have the same distribution on $(\mathcal G, J)$. Equivalently, the distributions of $[\bs G_1, \bs o_1]$ and $[\bs G_2,\bs o_2]$ agree on the invariant sigma-field $I$.

18: 	\end{description}

19: 	%	ater: update this. Condition~\ref{conditionB} is called \textbf{weak} and all other senses are called \textbf{strong}. This will justified by Theorem~\ref{thm:implications}.

20: \end{definition}

21: