1: \begin{abstract}

2:   We introduce  probability-graphons which are probability  kernels that

3:   generalize    graphons   to    the    case    of   weighted    graphs.

4:   Probability-graphons appear as the limit objects to study sequences of

5:   large  weighted   graphs  whose  distribution  of   subgraph  sampling

6:   converge.  The  edge-weights are  taken from  a general  Polish space,

7:   which also covers  the case of decorated graphs.  Here,  graphs can be

8:   either  directed or  undirected.   Starting from  a distance  $\dmeas$

9:   inducing the  weak topology on measures,  we define a cut  distance on

10:   probability-graphons,  making  it  a   Polish  space,  and  study  the

11:   properties  of  this  cut  distance.   In  particular,  we  exhibit  a

12:   tightness  criterion  for  probability-graphons  related  to  relative

13:   compactness  in the  cut  distance.   We also  prove  that under  some

14:   conditions  on the  distance $\dmeas$,  which are  satisfied for  some

15:   well-know distances like the Prohorov distance, and the Fortet-Mourier

16:   and  Kantorovitch-Rubinstein norms,  the topology  induced by  the cut

17:   distance on the space of  probability-graphons is independent from the

18:   choice of $\dmeas$.  Eventually, we prove that this topology coincides

19:   with the  topology induced by  the convergence in distribution  of the

20:   sampled subgraphs.

21: \end{abstract}

22: