Session 27. Probabilistic and Extremal Combinatorics

Graph limits via discrete potential theory

Rudolf Grübel, Leibniz Universität Hannover, Germany
We present a probabilistic approach to the currently very active area of graph limits; see [3] for background and references. Many of the models considered in this area can be regarded as combinatorial Markov chains \(X=(X_n)_{n\in\mathbf{N}}\) that are adapted to the family \(\mathcal{G}\) of finite simple graphs in the sense that \(X_n\) takes its values in the subset \(\mathcal{G}_n\) of graphs with \(n\) nodes. For stochastic processes of this kind discrete potential theory, as initiated by Doob in [1], provides a state space compactification with the properties that \(X_n\) converges to some \(X_\infty\) with values in the boundary \(\partial\mathcal{G}\), and that fully captures the persisting randomness of the process in the sense that the limit \(X_\infty\) generates the tail \(\sigma\)-field associated with \(X\). We review these concepts and then apply them to several popular graph models.
References
  1. J. Doob, Discrete potential theory and boundaries , J. Math. Mech. 8, 1959, 433-458.
  2. R. Grübel, Persisting randomness in randomly growing discrete structures: graphs and search trees , submitted, 2014.
  3. L. Lovász, Large networks and graph limits , American Mathematical Society Colloquium Publications, vol. 60, American Mathematical Society, Providence, RI 2012.
Print version