Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
184 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Graph limits and exchangeable random graphs (0712.2749v1)

Published 17 Dec 2007 in math.PR and math.CO

Abstract: We develop a clear connection between deFinetti's theorem for exchangeable arrays (work of Aldous--Hoover--Kallenberg) and the emerging area of graph limits (work of Lovasz and many coauthors). Along the way, we translate the graph theory into more classical probability.

Citations (388)

Summary

  • The paper formally connects exchangeable random graphs, derived from de Finetti-type theorems, with graph limit theory, characterizing infinite exchangeable graphs as distributions over graph limits.
  • The Aldous P Hoover theorem establishes a key representation for exchangeable arrays using a symmetric function \( W(x,y) \), which bridges probabilistic and combinatorial views and enables approximating large graphs with continuous objects.
  • These theoretical connections have profound implications for areas like statistical network analysis, randomized algorithms, and the study of complex systems, with potential future extensions to sparse graphs.

Graph Limits and Exchangeable Random Graphs

The paper "Graph Limits and Exchangeable Random Graphs" by Persi Diaconis and Svante Janson presents a detailed analysis of the connection between de Finetti's theorem and the theory of graph limits. The work lies at the intersection of probability theory and combinatorial graph theory, providing insights into understanding large graph structures.

Exchangeable Arrays and Their Extensions

The paper begins by examining the foundational aspects of de Finetti's theorem for exchangeable arrays, with extensions developed by Aldous, Hoover, and Kallenberg. An exchangeable array involves a sequence of random variables that exhibit a specific symmetry property under permutations. For binary sequences, de Finetti's theorem enables their representation as mixtures of independent and identically distributed (i.i.d.) sequences.

Diaconis and Janson extend this framework to two-dimensional arrays. The Aldous–Hoover theorem establishes a complex representation for separately and jointly exchangeable arrays, expressed via a function W:[0,1][0,1]W: [0,1] \to [0,1] that can be modeled by mixing such functions. This representation plays a pivotal role in translating between probabilistic and combinatorial interpretations.

Graph Limit Theory

The authors refer to the burgeoning field of graph limits, particularly the work of Lovász and his collaborators, which provides a foundational theory for understanding sequences of large graphs. The core concept is the convergence of a sequence of graphs, defined via homomorphism densities. In dense graphs, this is analyzed through graph homomorphisms, deriving limits that provide approximate properties of large, complex graph structures.

The relationship between exchangeable random graphs and graph limits is illuminated through the concept of a "limit object," a notion capturing the limiting behavior of a sequence of finite graphs or graphons. Here, function W(x,y)W(x,y), symmetric in nature, underlies the approximation of large graphs by continuous objects, providing a bridge to algorithmic applications, such as graph property testing and parameter estimation.

Main Results and Implications

A significant contribution of the paper is the formalization of the connection between exchangeable random graphs and the theory of graph limits. This involves characterizing infinite exchangeable random graphs as distributions over a space of graph limits. By utilizing the extreme point characterization of exchangeable distributions, the authors establish a one-to-one correspondence between exchangeable random graph limits and deterministic graph limits.

The extension to bipartite and directed graphs further expands the scope of the results, employing similar methodologies to derive appropriate probabilistic representations for these graph classes.

The theoretical advancements presented have profound implications for both mathematics and applied domains, particularly in areas such as statistical network analysis, randomized algorithm design, and the paper of complex systems. Future research could explore the applicability of these results to sparse graphs, an area where initial results are beginning to emerge.

Conclusion

"Graph Limits and Exchangeable Random Graphs" provides a comprehensive account of the intricate connections between graph theory and probability theory. By bridging de Finetti-type probabilistic symmetries with graph limit frameworks, Diaconis and Janson offer a structured approach to analyzing large-scale graph structures, fostering further advancements in both theoretical exploration and practical implementations in various scientific fields.