--- title: Edge-Measure Matrices url: https://www.emergentmind.com/topics/edge-measure-matrices type: topic --- # Edge-Measure Matrices Searching arXiv for the supplied papers and closely related context. Edge-measure matrices are matrix or operator realizations of edge-based weights, distances, flows, or boundary measurements. Across several distinct literatures, the phrase can refer to at least four mathematically different constructions: a diagonal operator attached to the equivariant edge measure on partitions in Donaldson–Thomas theory and Jack combinatorics; Gram- and Laplacian-type matrices built from cycle spaces of graphs; vertex–edge distance matrices used in edge metric dimension; and boundary measurement matrices of directed networks on surfaces. In each case, the matrix packages edge-level combinatorial or geometric data into a linear-algebraic object whose entries, spectrum, minors, or induced measures encode structural information [2410.03912], [2305.13569], [2003.04045], [1609.07525], [1908.07437]. ## 1. Equivariant edge measure on partitions and its matrix realization In the setting of Donaldson–Thomas invariants of toric threefolds, the equivariant edge measure is a probability measure on partitions arising from the edge term in the virtual localization formula. For a partition $\lambda$, identified with its Young diagram in English notation with zero-indexed coordinates, $$ \lambda=\{(i,j)\mid 0\le i\le \text{len}(\lambda)-1,\ 0\le j\le \lambda_i-1\}, $$ one introduces $$ Q(\lambda) = \sum_{(i,j)\in\lambda} r^i s^j,\qquad \overline{Q}(\lambda) = \sum_{(i,j)\in\lambda} r^{-i} s^{-j}, $$ and then defines $$ F(\lambda) = - Q(\lambda) - \frac{\overline{Q}(\lambda)}{rs} + \frac{Q(\lambda)\,\overline{Q}(\lambda)\,(1-r)(1-s)}{rs}. $$ Writing $F(\lambda)=\sum c_{ij}r^is^j$, the swap operation is $$ \text{swap}(G) = \prod_{(i,j)} (i u - j v)^{c_{ij}}, $$ for Laurent polynomials with no constant term, and the equivariant edge weight is $$ w_{\text{edge}}(\lambda):=\text{swap}\bigl(F(\lambda)\bigr). $$ Normalizing these weights gives a probability measure on partitions whenever the normalization sum makes sense [2410.03912]. The geometric origin is the edge contribution in Atiyah–Bott localization for Donaldson–Thomas theory of a toric $3$-fold with torus action. The fixed-point data factorize into vertex contributions indexed by $3$-dimensional partitions and edge contributions indexed by $2$-dimensional partitions, and the equivariant edge measure is the multiplicative Euler-class form of the Laurent polynomial $F(\lambda)$ attached to an edge partition [2410.03912]. The matrix interpretation is explicit in the paper’s overview. The weight function $w(\lambda)$ can be viewed as a diagonal operator on the partition basis, and via Jack polynomials it can also be realized as entries in Gram matrices, transition kernels, or projection operators in a Fock/Jacobi–Jack space. In the partition basis $\{\ket{\lambda}\}$, the diagonal operator is $$ \mathbf{W}_{\text{edge}}\ket{\lambda}=w_{\text{edge}}(\lambda)\ket{\lambda}, $$ so its matrix is $$ (\mathbf{W}_{\text{edge}})_{\lambda,\mu}=w_{\text{edge}}(\lambda)\,\delta_{\lambda\mu}. $$ This is the most direct sense in which the paper frames “edge-measure matrices”: matrices whose diagonal entries are the equivariant edge weights [2410.03912]. A key combinatorial tool is the corner polynomial $$ C(\lambda):=Q(\lambda)(1-r)(1-s), $$ whose expression in terms of inside and outside corners enables ratio computations when one box is added or removed. Another basic identity is $$ \text{swap}(F-G)=\frac{\text{swap}(F)}{\text{swap}(G)}, $$ which underlies multiplicative comparison of weights [2410.03912]. ## 2. Identification with the Jack–Plancherel measure The main result of "Jack combinatorics of the equivariant edge measure" is that the equivariant edge measure coincides, up to a global sign and normalization conventions, with the Jack–Plancherel measure [2410.03912]. For a box $\square=(i,j)\in\lambda$, the paper uses homogenized upper and lower Jack hook lengths $$ h^*_\lambda(\square)=u\,\ell_\lambda(\square)+v\bigl(a_\lambda(\square)+1\bigr),\qquad h_*^\lambda(\square)=u\bigl(1+\ell_\lambda(\square)\bigr)+v\,a_\lambda(\square), $$ where $a_\lambda(\square)$ and $\ell_\lambda(\square)$ are the arm and leg lengths. The Jack–Plancherel weight is then $$ w_{\text{Jack}}(\lambda) := \frac{1}{\displaystyle\prod_{\square\in\lambda} h^*_\lambda(\square)\,h_*^\lambda(\square)}, $$ with Jack parameter $\alpha=v/u$ [2410.03912]. The theorem is stated as $$ w_{\text{Jack}}(\lambda)=-\,w_{\text{edge}}(\lambda) $$ for all partitions $\lambda$. After normalization, the two probability measures coincide. The proof is ratio-based rather than termwise: the two weights are shown to have the same multiplicative growth rule under addition of a single box, and then an induction anchored at the one-box partition yields the identity [2410.03912]. This identification has immediate matrix consequences. The edge-measure diagonal operator and the corresponding Jack operator differ only by a global sign, so the edge-measure matrix may equally be interpreted as the inverse diagonal Gram form for Jack polynomials. In the Jack basis $\{\mathrm P_\lambda^{(\alpha)}\}$ with $$ \langle \mathrm{P}_\lambda^{(\alpha)},\mathrm{P}_\mu^{(\alpha)}\rangle_\alpha = \delta_{\lambda\mu}\, j_\lambda^{(\alpha)}, $$ the diagonal Gram matrix has entries $G_{\lambda\lambda}=j_\lambda^{(\alpha)}$, while the Jack–Plancherel weight is proportional to $1/j_\lambda^{(\alpha)}$. This means the edge measure can be read as the diagonal of the inverse Gram matrix, or as eigenvalues of a projection operator normalized by Jack norms [2410.03912]. The paper also describes a transition-matrix viewpoint on Young’s lattice. For partitions $\mu\subset\lambda$ with $|\lambda|=|\mu|+1$, one may define transition probabilities proportional to $$ R(\mu\to\lambda):=\frac{w(\lambda)}{w(\mu)}, $$ and the explicit corner formulas provide closed forms for every nonzero entry of the transition matrix $T_n$. This suggests that edge-measure matrices can also be regarded as transition kernels for growth processes whose stationary or terminal distributions are Jack–Plancherel distributions [2410.03912]. ## 3. Mesh matrices and edge-based Gram forms on graphs A distinct meaning of edge-measure matrix arises in graph theory through the mesh matrix $Mesh(G,T_0)$ of a connected finite graph $G$ relative to a spanning tree $T_0$. Given the non-tree edges $e_1,\dots,e_N$ and the associated fundamental cycles $$ Z[j]=1\cdot e_j + D(1\cdot e_j)\in C_1(G;\mathbb R), $$ the mesh matrix is the Gram matrix $$ Mesh(G,T_0)=\big(\langle Z[j_1],Z[j_2]\rangle\big)_{1\le j_1,j_2\le N} $$ in the standard inner product on the edge space [2305.13569]. This is explicitly an edge-based matrix. The basis vectors $Z[j]$ are supported on edges, each consists of a non-tree edge plus the unique path in the spanning tree joining its endpoints, and the resulting matrix records pairwise overlap of these edge-supported cycles. Writing the matrix of the cycle basis in the full edge basis as $$ Y[+]=\begin{pmatrix}I_N\\ Y\end{pmatrix}, $$ one obtains $$ Mesh(G,T_0)=Y[+]^\top Y[+]=I_N+Y^\top Y. $$ Thus the reduced mesh matrix is $$ Mesh^\#(G,T_0):=Y^\top Y, $$ and the full mesh matrix has all eigenvalues real and $\ge 1$ because it is of the form $I_N+A^\top A$ [2305.13569]. The paper gives an “edge-measure interpretation of the spectrum.” For $u\in\mathbb R^N$, the vector $Yu$ is a $1$-chain on the tree, representing the superposition of tree paths weighted by $u_j$, and $$ u^\top Mesh^\#(G,T_0)u=\|Yu\|^2 $$ is the total squared flow through tree edges induced by that linear combination of cycles. Large eigenvalues correspond to directions in cycle space forcing large total flow through the tree; small eigenvalues correspond to near-cancellations [2305.13569]. The characteristic polynomial has a spanning-tree interpretation: $$ \det(XI-Mesh(G,T_0)) = \sum_{j=0}^{N}(-1)^j ST_j(G,T_0)(X-1)^{N-j}, $$ where $ST_j(G,T_0)$ counts spanning trees meeting the non-tree edge set in exactly $j$ edges. The matrix therefore refines spanning-tree statistics by edge usage. This suggests an edge-measure viewpoint in which the spectrum records how spanning trees distribute their usage of off-tree edges [2305.13569]. The associated mesh Laplacian $$ L_{mesh}(G,T_0)=YY^\top=DD^\ast $$ is a generalization of the Kirchhoff Laplacian. In a coned-graph construction it reduces exactly to the standard graph Laplacian $Deg-Adj$, and its smallest positive eigenvalue is interpreted as a measure of flux from the tree through the complement $G-T_0$ [2305.13569]. ## 4. Edge–vertex distance matrices and edge metric dimension In metric graph theory, the phrase “edge-measure matrix” is naturally attached to the edge–vertex distance matrix used to study edge metric dimension. For a connected graph $G$ with vertices $V(G)=\{v_1,\dots,v_n\}$ and edges $E(G)=\{e_1,\dots,e_m\}$, the distance from a vertex to an edge is $$ d_G(u,xy)=\min\{d_G(u,x),d_G(u,y)\}, $$ and the edge metric representation of an edge $e$ with respect to an ordered set $S=\{v_{j_1},\dots,v_{j_k}\}$ is $$ r_G(e\mid S)=\bigl(d_G(e,v_{j_1}),\dots,d_G(e,v_{j_k})\bigr). $$ The paper introduces the full matrix $$ D_G=[d_{ij}],\qquad d_{ij}=d_G(e_i,v_j), $$ whose rows correspond to edges and columns to vertices [2003.04045]. This matrix is not symmetric and is rectangular of size $m\times n$. Selecting a subset of columns indexed by $S$ produces the restricted matrix whose $i$th row is exactly $r_G(e_i\mid S)$. A set $S$ is an edge metric generator precisely when all rows of that submatrix are pairwise distinct. In this sense, edge metric dimension is the minimum number of columns required so that the edge-measure matrix has distinct rows [2003.04045]. The paper gives an integer linear programming formulation directly in terms of $D_G$. With binary variables $x_i\in\{0,1\}$ indicating whether $v_i$ is selected, one minimizes $$ x_1+\cdots+x_n $$ subject to $$ |d_{i1}-d_{j1}|x_1+\cdots+|d_{in}-d_{jn}|x_n>0 \qquad (1\le i