Core Concepts
This paper presents two lower bounds for the intersection number of an arbitrary connected graph, which is a measure of the sparsity of the cycle intersection matrix induced by a spanning tree. The first lower bound is proven, while the second is conjectured based on experimental results.
Abstract
The paper focuses on the Minimum Spanning Tree Cycle Intersection (MSTCI) problem, which aims to find a spanning tree that minimizes the number of non-empty pairwise intersections of its tree-cycles. This problem is relevant for integrating discrete differential forms on graphs.
The key points are:
The authors prove a lower bound for the intersection number of a connected graph G = (V, E):
1/2 * (ν^2 / (n-1) - ν) ≤ ∩(G)
where n = |V| and ν is the cyclomatic number of G.
Based on experimental results and a new observation, the authors conjecture an improved tight lower bound:
(n-1) * √(q/2) + q*r ≤ ∩(G)
where 2ν = q(n-1) + r is the integer division of 2ν and n-1.
The authors discuss the significance of these lower bounds in approximating the sparsity of cycle intersection matrices, which is important for applying fast linear solvers to the mesh deformation problem.
The authors also analyze the case of graphs with a universal vertex, showing that the star spanning tree is a solution to the MSTCI problem for such graphs.
The paper provides several auxiliary lemmas and definitions related to bonds, non-redundant bond sets, and the properties of the cycle intersection matrix.