Lower bounds on the distortion of embedding finite metric spaces in graphs (Q1380790)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Lower bounds on the distortion of embedding finite metric spaces in graphs
scientific article

    Statements

    Lower bounds on the distortion of embedding finite metric spaces in graphs (English)
    0 references
    0 references
    22 June 1998
    0 references
    Given two finite metric spaces \((X,d)\) and \((Y,\delta)\) with \(|X|=|Y|\), and a \(1-1\) map \(f:X \to Y\), the Lipschitz norm \(||f||\) of \(f\) is \text{\(\displaystyle\max_{s\neq t \in X}{\delta(f(s),f(t))\over d(s,t)}\)} and the Lipschitz distance between \(X\) and \(Y\) is \(\text{dist}(X,Y)=\inf_{f:X\to Y} ||f||\cdot ||f^{-1}||\), where the infimum is taken over all \(1-1\) maps. The authors investigate the question of how well a finite metric space can be represented by a subspace of a graph, subject to certain restrictions. Specifically, they prove (1) if \(H\) is a simple, unweighted connected graph with \(n\) vertices, and if \(G\) is any (weighted) graph with the same number of vertices but strictly fewer edges than \(H\), then \(\text{dist}(H,G)\geq g/3 - 1\), where \(g\) is the girth of \(H\), and (2) with \(H\) as above, if \(G\) is any weighted graph with at least \(n\) vertices, and with \(\chi(G)<\chi(H)\), then for any \(n\)-point subspace \(S\) of the vertices of \(G\), \(\text{dist}(H,S)\geq g/4 - {3\over 2}\). They also investigate how well \(H\) can be approximated by a subspace of a graph \(G\) with \(\chi(G)<\chi(H)-1\) (in this case, the minimum distance is a function of the length of a second shortest cycle of \(H\)) and of how well an \(n\)-circuit can be approximated by a subspace of a tree (the distance is bounded below by \(n/3 -1\)). The latter result is then generalized to the following question: how well can the \(n\)-sphere be approximated by a subspace of a simplicial complex \(K\) of dimension \(\leq n\) with \(H_n(K)\) trivial?
    0 references
    0 references
    finite metric space
    0 references
    graph embedding
    0 references
    0 references
    0 references
    0 references