arXiv++ Combinatorics

For a given nonnegative matrix $A=(A_{ij})$, the matrix scaling problem asks whether $A$ can be scaled to a doubly stochastic matrix $D_1AD_2$ for some positive diagonal matrices $D_1,D_2$.The Sinkhorn algorithm is a simple iterative algorithm, which repeats row-normalization $A_{ij} \leftarrow A_{ij}/\sum_{j}A_{ij}$ and column-normalization $A_{ij} \leftarrow A_{ij}/\sum_{i}A_{ij}$ alternatively. By this algorithm, $A$ converges to a doubly stochastic matrix in limit if and only if the bipartite graph associated with $A$ has a perfect matching. This property can decide the existence of a perfect matching in a given bipartite graph $G$, which is identified with the $0,1$-matrix $A_G$.Linial, Samorodnitsky, and Wigderson showed that $O(n^2 \log n)$ iterations for $A_G$ decide whether $G$ has a perfect matching. Here $n$ is the number of vertices in one of the color classes of $G$. In this paper, we show an extension of this result:If $G$ has no perfect matching, then a polynomial number of the Sinkhorn iterations identifies a Hall blocker -- a vertex subset $X$ having neighbors $Γ(X)$ with $|X| > |Γ(X)|$. Specifically, we show that $O(n^2 \log n)$ iterations can identify one Hall blocker, and that further polynomial iterations can also identify all parametric Hall blockers $X$ of maximizing $(1-λ) |X| - λ|Γ(X)|$ for $λ\in [0,1]$.The former result is based on an interpretation of the Sinkhorn algorithm as alternating minimization for geometric programming. The latter is on an interpretation as alternating minimization for KL-divergence (Csiszár and Tusnády 1984, Gietl and Reffel 2013) and its limiting behavior for a nonscalable matrix (Aas 2014). We also relate the Sinkhorn limit with parametric network flow, principal partition of polymatroids, and the Dulmage-Mendelsohn decomposition of a bipartite graph.

A basic property in a modular lattice is that any two flags generate a distributive sublattice. It is shown (Abels 1991, Herscovic 1998) that two flags in a semimodular lattice no longer generate such a good sublattice, whereas shortest galleries connecting them form a relatively good join-sublattice. In this note, we sharpen this investigation to establish an analogue of the two-flag generation theorem for a semimodular lattice. We consider the notion of a modular convex subset, which is a subset closed under the join and meet only for modular pairs, and show that the modular convex hull of two flags in a semimodular lattice of rank $n$ is isomorphic to a union-closed family on $[n]$. This family uniquely determines an antimatroid, which coincides with the join-sublattice of shortest galleries of the two flags.