1: \begin{proof}[Proof of Lemma \ref{lem:analysis_formula}]

2: 	

3: 	Indeed, a more rigorous expression of Equation (\ref{eq:analysis_formula}) is

4: 	\begin{equation}\label{eq:analysis_formula_rigorous}

5: 	\begin{aligned}

6: 	P_v(S_m)=\sum_{S_{m-1}\subseteq V_{m-1}}\prod_{u\in S_{m-1}}f_u(S_{m})\prod_{u\in V_{m-1},~u\notin S_{m-1}}(1-f_u(S_m))P_v(S_{m-1}).

7: 	\end{aligned}

8: 	\end{equation}

9: 	

10: 	However, for any $u\notin V_{m-1}$, $f_u(S_m)=0$, this implies that (\ref{eq:analysis_formula}) and (\ref{eq:analysis_formula_rigorous}) are equivalent.

11: 	Without loss of generality, we assume $v\in V_1$ and show Equation (\ref{eq:analysis_formula}).

12: 	

13: 	When $m=2$, $P_v(S_2)=f_v(S_2)$ satisfies Equation (\ref{eq:analysis_formula}).

14: 	When $m>2$,

15: 	let $P_{S_{m-1}}(S_m)$ denote the probability that the active node set in $V_{m-1}$ is exactly $S_{m-1}$ when the seed set is $S_{m}\subseteq V_{m}$.

16: 	Then, for any $S_{m-1}\subseteq V_{m-1}$ and a fix seed set $S_{m}\subseteq V_{m}$, we have $P_{S_{m-1}}(S_{m})=\prod_{u\in S_{m-1}}f_u(S_{m})\prod_{u\notin S_{m-1}}(1-f_u(S_{m}))$ since the threshold value of each node is generated independently.

17: 	

18: 	Given $S_m\subseteq V_m$, $S_{m-1}\subseteq V_{m-1}$ and $v\in V_1$, let $\mathcal{E}_1$ be the random event that the active node set in $V_{m-1}$ is exactly $S_{m-1}$ when the seed set is $S_{m}$ and $\mathcal{E}_2$ be the random event that $v$ can be activated when the active nodes set in $V_{m-1}$ is $S_{m-1}$.

19: 	It is obvious that $\mathcal{E}_1$ and $\mathcal{E}_2$ are two independent random events, thus,

20: 	\begin{equation*}

21: 	\begin{aligned}

22: 	P_v(S_{m})&=\sum_{S_{m-1}\subseteq V_{m-1}}P_{S_{m-1}}(S_m)P_v(S_{m-1})\\

23: 	&=\sum_{S_{m-1}\subseteq V_{m-1}}\prod_{u\in S_{m-1}}f_u(S_{m})\prod_{u\notin S_{m-1}}(1-f_u(S_{m}))P_v(S_{m-1}).

24: 	\end{aligned}

25: 	\end{equation*}

26: \end{proof}

27: