1: \begin{abstract} %

2: In a recent paper, Caron and Fox suggest a probabilistic model for sparse

3: graphs which are exchangeable when associating each vertex with a time

4: parameter in $\R_+$. Here we show that by generalizing the classical

5: definition of graphons as functions over probability spaces to functions

6: over $\sigma$-finite measure spaces, we can model a large family of

7: exchangeable graphs, including the Caron-Fox graphs and the traditional

8: exchangeable dense graphs as special cases. Explicitly, modelling the

9: underlying space of features by a $\sigma$-finite measure space

10: $(S,\cS,\mu)$ and the connection probabilities by an integrable function

11: $W\colon S\times S\to [0,1]$, we construct a random family  $(G_t)_{t\geq

12: 0}$ of growing graphs such that the vertices of $G_t$ are given by a

13: Poisson point process on $S$ with intensity $t\mu$, with two points $x,y$

14: of the point process connected with probability $W(x,y)$. We call such a

15: random family a \emph{graphon process}. We prove that a graphon process has

16: convergent subgraph frequencies (with possibly infinite limits) and that,

17: in the natural extension of the cut metric to our setting, the sequence

18: converges to the generating graphon. We also show that the underlying

19: graphon is identifiable only as an equivalence class over graphons with cut

20: distance zero. More generally, we study metric convergence for arbitrary

21: (not necessarily random) sequences of graphs, and show that a sequence of

22: graphs has a convergent subsequence if and only if it has a subsequence

23: satisfying a property we call \emph{uniform regularity of tails}. Finally,

24: we prove that every graphon is equivalent to a graphon on $\R_+$ equipped

25: with Lebesgue measure.

26: \end{abstract}

27: