On the relation between weighted trees and tropical Grassmannians (Q1023271): Difference between revisions

If \(T\) is an \(n\)--leaved metric tree and \(D(i,j)\) is the distance between leaf \(i\) and leaf \(j\), then the zero diagonal, symmetric real matrix \(D\) is called the dissimilarity matrix associated to \(T\). A theorem of \textit{P. Buneman} [J. Comb. Theory, Ser. B 17, 48--50 (1974; Zbl 0286.05102)] characterizes dissimilarity matrices as those zero diagonal, symmetric real matrices satisfying the \textit{four--point condition} as follows: for each \(i,j,k,l\in\{1,\ldots,n\}\) not necessarily different, the maximum \(\max\{D(i,j)+D(k,l), D(i,k)+D(j,l), D(i,l)+D(j,k)\}\) is attained, at least, twice. In today's language, this is a tropical vanishing condition! The notion of dissimilarity matrix (\(m=2\)) is generalized to that of family of \(m\)--dissimilarity vectors (for \(2\leq m<n\)). The main theorem of the paper proves that the family of \(4\)--dissimilarity vectors is contained in the tropical Grassmannian \(\mathcal G_{4,n}\). This follows the trace of a theorem of \textit{D. Speyer} and \textit{B. Sturmfels} [Adv. Geom. 4, No. 3, 389--411 (2004; Zbl 1065.14071)] claiming that the family of \(2\)--dissimilarity vectors equals the tropical Grassmannian \(\mathcal G_{2,n}\) and a theorem of \textit{C. Bocci} and the author [Appl. Math. Comput. 212, No. 2, 349--356 (2009; Zbl 1220.05016)] relating the family of \(3\)--dissimilarity vectors to the tropical Grassmannian \(\mathcal G_{3,n}\). It is conjectured that the main theorem holds true for any \(m\geq5\). The conjecture has been checked out for \(m=5\), using a computer algebra system.