1: \begin{abstract}

2: We consider random networks whose dynamics is

3: described by a rate equation, with transition rates $w_{nm}$

4: that form a symmetric matrix. The long time evolution

5: of the system is characterized by a diffusion coefficient~$D$.

6: In one dimension it is well known that $D$ can display an abrupt

7: percolation-like transition from diffusion (${D>0}$)

8: to sub-diffusion (${D=0}$). A question arises whether

9: such a transition happens in higher dimensions.

10: Numerically $D$ can be evaluated using a resistor network

11: calculation, or optionally it can be deduced from 

12: the spectral properties of the system. Contrary to a recent 

13: expectation that is based on a renormalization-group analysis, 

14: we deduce that $D$ is finite; 

15: suggest an ``effective-range-hopping" procedure to evaluate it;

16: and contrast the results with the linear estimate.

17: The same approach is useful for the analysis of 

18: networks that are described by quasi-one-dimensional  

19: sparse banded matrices. 

20: \end{abstract}

21: