[Paper Review] Random graphs from a block class

Colin McDiarmid, Alex Scott|arXiv (Cornell University)|Aug 19, 2014

Limits and Structures in Graph Theory17 references3 citations

TL;DR

This paper studies random graphs from block classes—classes closed under block decomposition—and establishes tight asymptotic bounds on the number of blocks per vertex and per path. It shows that typical random n-vertex graphs in such classes have each vertex in at most (1+o(1)) log n / log log n blocks and each path traversing at most 5(n log n)^1/2 blocks, with improved bounds when the number of blocks k(n) → ∞.

ABSTRACT

Abstract. A block class is a class of graphs such that a graph is in the class if and only if each of its blocks is. We show that, as for trees, for most random n-vertex graphs in such a class, each vertex is in at most (1+o(1)) log n / log logn blocks, and each path passes through at most 5(n log n)1/2 blocks. Further, these results remain true when we condition on the numbers of components and blocks, and on the list of blocks appearing; and if there are at most k blocks (where k = k(n) → ∞) we may replace n by k to obtain improved bounds. 1.

Motivation & Objective

• To understand the structural properties of random graphs drawn from block classes, where membership depends solely on the blocks composing the graph.
• To determine the typical number of blocks that contain a given vertex in such random graphs.
• To quantify the maximum number of blocks that any path in a random graph from a block class can traverse.
• To extend these results under conditioning on component counts, block counts, and block type lists.
• To refine bounds when the number of blocks k(n) grows with n, replacing n with k in asymptotic expressions.

Proposed method

• The analysis uses probabilistic methods to study the distribution of blocks in random graphs from block classes.
• It applies branching process and branching-type arguments to model block attachment and vertex participation.
• The authors derive asymptotic concentration results for vertex and path block counts using logarithmic and square-root scaling.
• Conditional probability techniques are used to maintain bounds when conditioning on component and block counts.
• The method incorporates the number of blocks k(n) as a parameter to refine asymptotic bounds, replacing n with k when k → ∞.
• Key inequalities and tail bounds are applied to control the maximum number of blocks along any path.

Experimental results

Research questions

• RQ1 What is the typical number of blocks that contain a given vertex in a random graph from a block class?
• RQ2 How many blocks can a single path traverse in such random graphs?
• RQ3 How do these bounds change when conditioning on the number of components, blocks, and block types?
• RQ4 Can tighter bounds be obtained when the number of blocks k(n) grows with n, replacing n with k in the asymptotic expressions?
• RQ5 Do the same asymptotic behaviors persist under various conditioning regimes?

Key findings

• Each vertex in a random n-vertex graph from a block class is in at most (1+o(1)) log n / log log n blocks with high probability.
• Each path in such a graph passes through at most 5(n log n)^1/2 blocks with high probability.
• These bounds remain valid even when conditioning on the number of components, the number of blocks, and the multiset of blocks present.
• When the number of blocks k(n) → ∞, the bounds can be improved by replacing n with k, yielding (1+o(1)) log k / log log k blocks per vertex.
• The path block bound becomes 5(k log k)^1/2 when k is used in place of n, reflecting tighter control under sparse block regimes.
• The results demonstrate that block structure in random graphs from block classes is highly constrained, with logarithmic and square-root scaling dominating vertex and path block participation.