arXiv subjects

Varun Sivashankar

Publications and source records attributed to Varun Sivashankar.

16 recordsLinked to original sources

arXiv2026

Zero-Sum Cycles in Regular Digraphs

Let $\Gamma$ be a finite group of order $k\ge2$, and label the edges of a simple loopless $d$-regular digraph $D$ by elements of $\Gamma$. A directed cycle is zero-sum if the ordered product of its labels is the identity of $\Gamma$. We prove that a zero-sum cycle exists whenever $d\ge e^3(k-1)$. We also prove that every labelled $d$-regular digraph contains $\Omega(d/k)$ pairwise vertex-disjoint zero-sum cycles. When $d\ge50k$, it contains $\Omega(d^2/k)$ pairwise edge-disjoint zero-sum cycles. All three results are asymptotically optimal. The existence and packing results extend to Eulerian digraphs whose minimum and maximum common degrees $\delta$ and $\Delta$ satisfy $\delta^3/\Delta^2=\Omega(k)$. The techniques extend a determinant--permanent argument of Friedland for even directed cycles.

math.CO↗

arXiv2026

Graph Eigenvalues and Projection Constants

For an integer $k\ge2$, let $\lambda_k(G)$ denote the $k$th largest adjacency eigenvalue of a graph $G$. For every graph $G$ on $n$ vertices and every $2 \leq k \leq n$, we prove \[ \lambda_k(G) \le \frac{(k-2)\sqrt{k+1}+2}{2k(k-1)}\,n-1. \] Our bound is tight for $k\in\{2,3,4,8,24\}$. We obtain it by reducing the graph-eigenvalue problem to an extremal problem for orthogonal projections and then applying the general upper bound on the absolute projection constant $\gamma(r)$ due to Der\k{e}gowska and Lewandowska. We also give an alternative proof of their bound by repairing the Gegenbauer-polynomial argument of K\"onig and Tomczak-Jaegermann. The resulting slack identity yields a strict improvement in every even dimension $r\ge4$ for which $r+2$ is not a perfect square.

math.CO↗

arXiv2026

A Linear Bound on the Rainbow Cycle Number and Approximate EFX

It is open whether every fair-division instance with additive valuations admits a complete envy-free-up-to-any-good (EFX) allocation. A well-studied relaxation allows some goods to remain unallocated and asks for $(1-\varepsilon)$-EFX. The rainbow cycle number $R(d)$ was introduced to study this problem: upper bounds on $R(d)$ yield approximate EFX allocations with few unallocated goods. The best previous bound, $R(d)=O(d\log d)$, gives $O_\varepsilon(\sqrt{n\log n})$ unallocated goods. We resolve the conjecture that $R(d)$ is linear by proving $R(d)<ed$. It follows that every instance with $n$ agents admits a partial $(1-\varepsilon)$-EFX allocation with $O(\sqrt{n/\varepsilon})$ unallocated goods. This is the best possible asymptotic guarantee on the number of unallocated goods obtainable from the rainbow-cycle reduction. We also give a randomized algorithm that finds such an allocation in expected time polynomial in the input size and $1/\varepsilon$.

cs.GT↗

arXiv2026

An Improved Lower Bound for the Erd\H{o}s-Lov\'asz Cover Number Problem

Let $g(r)$ be the minimum number of edges in an $r$-uniform intersecting hypergraph with cover number $r$. Erd\H{o}s and Lov\'asz proved the lower bound $g(r)\ge 8r/3-3$. We first give a completely elementary proof that $g(r)\ge 3r-4$. We then build on the same approach and apply Kahn's small-codegree hypergraph edge-colouring theorem to improve this to $g(r)\ge ((41-\sqrt{19})/12-o(1))r\approx 3.053r$. In particular, this shows that $g(r)>3r$ for all sufficiently large $r$, addressing a question of Erd\H{o}s.

math.CO↗

arXiv2026

Pseudoshattering Pairs

For two vectors $x,y\in [b]^k$, consider the bipartite graph with two copies of $[b]$ in which $i$ on the left is joined to $j$ on the right if $(x_t,y_t)=(i,j)$ for some coordinate $t$. We study the largest size of a family $C\subseteq [b]^k$ such that, for every two distinct $x,y\in C$, this bipartite graph contains a cycle. We give a natural construction for such families and conjecture that it is optimal whenever $k$ is large relative to $b$. We prove an LYM-type upper bound that is asymptotically tight with respect to this construction, and is exact when $k$ is large and divisible by $b$. We then refine the argument using a circular ordering, obtaining the sharp full-support bound when $k\equiv -1\pmod b$. In the case $b=3$, we prove the exact general result when $k\equiv -1\pmod 3$ and $k$ is sufficiently large. The problem is motivated by the Daniely--Shalev-Shwartz dimension and the pseudocube formulation of a higher-alphabet Sauer-Shelah-Perles lemma.

math.CO↗

arXiv2026

Tree-independence number and forbidden induced subgraphs: excluding a $6$-vertex path and a $(2,t)$-biclique

We show that for every positive integer ${t \geq 2}$ there exists an integer $s$ such that every graph that contains no induced subgraph isomorphic to either the $6$-vertex path or the $(2,t)$-biclique, the complete bipartite graph $K_{2,t}$, has tree-independence number at most $s$. This result makes partial progress on a conjecture of Dallard, Krnc, Kwon, Milani\v{c}, Munaro, \v{S}torgel, and Wiederrecht.

math.CO↗

arXiv2026

Upper bound on the $k$-th eigenvalue of a graph

We prove a general upper bound on the $k$-th adjacency eigenvalue of a graph. For $k\ge 2$, we show that \[ \lambda_k(G)\le \frac{(k-2)\sqrt{k+1}+2}{2k(k-1)}\,n-1 \] for every graph $G$ on $n$ vertices. We build on a recent approach that addresses the case $k=3$ and generalize the upper bound for all $k \geq 3$ by using the positivity of Gegenbauer polynomials. The upper bound is tight for $k \in \{2,3,4,8,24\}$. We also highlight the close relation of $\lambda_k(G)$ to questions about equiangular lines.

math.CO↗

arXiv2026

Extremal Graphs for the Lights Out Problem

Lights Out is a game played on a graph $G$ where every vertex has a light bulb that is either on or off, and pressing a vertex $v$ toggles the state of every vertex in the closed neighborhood of $v$. The goal is to find a subset of vertices $S$ such that pressing every vertex in $S$ results in all light bulbs being turned off. We study the extremal graphs for which pressing every vertex is the unique solution to the lights out problem given an initial configuration of all lights on. We show that a graph is extremal if and only if it is even and has an odd number of matchings. Furthermore, there is a bijection between the set of labeled $n$-vertex extremal graphs and the set of symmetric invertible matrices of size $n-2$ over $\mathbb{F}_2$. We prove that any even graph with no cycle of length $0\pmod 3$ must be extremal. We also demonstrate operations that build larger extremal graphs from smaller ones. Along the way, we prove using the polynomial method that in any even graph, the number of matchings of a fixed size covering an odd subset of vertices is even.

math.CO↗

arXiv2025

On a Ramsey--Turán variant of Roth's theorem

A classical theorem of Roth states that the maximum size of a solution-free set of a homogeneous linear equation $\mathcal{L}$ in $\mathbb{F}_p$ is $o(p)$ if and only if the sum of the coefficients of $\mathcal{L}$ is $0$. In this paper, we prove a Ramsey--Turán variant of Roth's theorem, with respect to a natural notion of ``structured'' sets introduced by Erdős and Sárközy in the 1970's. Namely, we show that the following statements are equivalent: $(a)$ Every solution-free set $A$ of $\mathcal{L}$ in $\mathbb{F}_p$ with $α(\mathrm{Cay}_{\mathbb{F}_p}(A)) = o(p)$ has size $o(p)$. $(b)$ There exists a non-empty \emph{subset} of coefficients of $\mathcal{L}$ with zero sum.

math.CO↗

arXiv2025

Intersecting hypergraphs with large cover number

In their famous 1974 paper introducing the local lemma, Erdős and Lovász posed a question-later referred by Erdős as one of his three favorite open problems: What is the minimum number of edges in an $r$-uniform, intersecting hypergraph with cover number $r$? This question was solved up to a constant factor in Kahn's remarkable 1994 paper. More recently, motivated by applications to Bollobás' ''power of many colours'' problem, Alon, Bucić, Christoph, and Krivelevich introduced a natural generalization by imposing a space constraint that limits the hypergraph to use only $n$ vertices. In this note we settle this question asymptotically, up to a logarithmic factor in $n/r$ in the exponent, for the entire range.

math.CO↗

arXiv2025

Tree independence number V. Walls and claws

Given a family $\mathcal{H}$ of graphs, we say that a graph $G$ is $\mathcal{H}$-free if no induced subgraph of $G$ is isomorphic to a member of $\mathcal{H}$. Let $S_{t,t,t}$ be the graph obtained from $K_{1,3}$ by subdividing each edge $t-1$ times, and let $W_{t\times t}$ be the $t$-by-$t$ hexagonal grid. Let $\mathcal{L}_t$ be the family of all graphs $G$ such that $G$ is the line graph of some subdivision of $W_{t \times t}$. We prove that for every positive integer $t$ there exists $c(t)$ such that every $\mathcal{L}_t \cup \{S_{t,t,t}, K_{t,t}\}$-free $n$-vertex graph admits a tree decomposition in which the maximum size of an independent set in each bag is at most $c(t)\log^4n$. This is a variant of a conjecture of Dallard, Krnc, Kwon, Milanič, Munaro, Štorgel, and Wiederrecht from 2024. This implies that the Maximum Weight Independent Set problem, as well as many other natural algorithmic problems, that are known to be NP-hard in general, can be solved in quasi-polynomial time if the input graph is $\mathcal{L}_t \cup \{S_{t,t,t},K_{t,t}\}$-free. As part of our proof, we show that for every positive integer $t$ there exists an integer $d$ such that every $\mathcal{L}_t \cup \{S_{t,t,t}\}$-free graph admits a balanced separator that is contained in the neighborhood of at most $d$ vertices.

math.CO↗

arXiv2024

Maximum shattering

A family $\mathcal{F}$ of subsets of $[n]=\{1,2,\ldots,n\}$ shatters a set $A \subseteq [n]$ if for every $A' \subseteq A$ there is an $F \in \mathcal{F}$ such that $F \cap A=A'$. We develop a framework to analyze $f(n,k,d)$, the maximum possible number of subsets of $[n]$ of size $d$ that can be shattered by a family of size $k$. Among other results, we determine $f(n,k,d)$ exactly for $d \leq 2$ and show that if $d$ and $n$ grow, with both $d$ and $n-d$ tending to infinity, then, for any $k$ satisfying $2^d \leq k \leq (1+o(1))2^d$, we have $f(n,k,d)=(1+o(1))c\binom{n}{d}$, where $c$, roughly $0.289$, is the probability that a large square matrix over $\mathbb{F}_2$ is invertible. This latter result extends work of Das and Mészáros. As an application, we improve bounds for the existence of covering arrays for certain alphabet sizes.

math.CO↗

arXiv2023

Composable Coresets for Determinant Maximization: Greedy is Almost Optimal

Given a set of $n$ vectors in $\mathbb{R}^d$, the goal of the \emph{determinant maximization} problem is to pick $k$ vectors with the maximum volume. Determinant maximization is the MAP-inference task for determinantal point processes (DPP) and has recently received considerable attention for modeling diversity. As most applications for the problem use large amounts of data, this problem has been studied in the relevant \textit{composable coreset} setting. In particular, [Indyk-Mahabadi-OveisGharan-Rezaei--SODA'20, ICML'19] showed that one can get composable coresets with optimal approximation factor of $\tilde O(k)^k$ for the problem, and that a local search algorithm achieves an almost optimal approximation guarantee of $O(k)^{2k}$. In this work, we show that the widely-used Greedy algorithm also provides composable coresets with an almost optimal approximation factor of $O(k)^{3k}$, which improves over the previously known guarantee of $C^{k^2}$, and supports the prior experimental results showing the practicality of the greedy algorithm as a coreset. Our main result follows by showing a local optimality property for Greedy: swapping a single point from the greedy solution with a vector that was not picked by the greedy algorithm can increase the volume by a factor of at most $(1+\sqrt{k})$. This is tight up to the additive constant $1$. Finally, our experiments show that the local optimality of the greedy algorithm is even lower than the theoretical bound on real data sets.

cs.DS↗

arXiv2023

Improved Approximation Algorithms for the Joint Replenishment Problem with Outliers, and with Fairness Constraints

The joint replenishment problem (JRP) is a classical inventory management problem. We consider a natural generalization with outliers, where we are allowed to reject (that is, not service) a subset of demand points. In this paper, we are motivated by issues of fairness - if we do not serve all of the demands, we wish to ``spread out the pain'' in a balanced way among customers, communities, or any specified market segmentation. One approach is to constrain the rejections allowed, and to have separate bounds for each given customer. In our most general setting, we consider a set of C features, where each demand point has an associated rejection cost for each feature, and we have a given bound on the allowed rejection cost incurred in total for each feature. This generalizes a model of fairness introduced in earlier work on the Colorful k-Center problem in which (analogously) each demand point has a given color, and we bound the number of rejections of each color class. We give the first constant approximation algorithms for the fairness-constrained JRP with a constant number of features; specifically, we give a 2.86-approximation algorithm in this case. Even for the special case in which we bound the total (weighted) number of outliers, this performance guarantee improves upon bounds previously known for this case. Our approach is an LP-based algorithm that splits the instance into two subinstances. One is solved by a novel iterative rounding approach and the other by pipage-based rounding. The standard LP relaxation has an unbounded integrality gap, and hence another key element of our algorithm is to strengthen the relaxation by correctly guessing key attributes of the optimal solution, which are sufficiently concise, so that we can enumerate over all possible guesses in polynomial time - albeit exponential in C, the number of features.

cs.DS↗

arXiv2022

Extremal Uniquely Resolvable Multisets

For positive integers $n$ and $m$, consider a multiset of non-empty subsets of $[m]$ such that there is a \textit{unique} partition of these subsets into $n$ partitions of $[m]$. We study the maximum possible size $g(n,m)$ of such a multiset. We focus on the regime $n \leq 2^{m-1}-1$ and show that $g(n,m) \geq Ω(\frac{nm}{\log_2 n})$. When $n = 2^{cm}$ for any $c \in (0,1)$, this lower bound simplifies to $Ω(\frac{n}{c})$, and we show a matching upper bound $g(n,m) \leq O(\frac{n}{c}\log_2(\frac{1}{c}))$ that is optimal up to a factor of $\log_2(\frac{1}{c})$. We also compute $g(n,m)$ exactly when $n \geq 2^{m-1} - O(2^{\frac{m}{2}})$.

math.CO↗

arXiv2021

Local Linearity and Double Descent in Catastrophic Overfitting

Catastrophic overfitting is a phenomenon observed during Adversarial Training (AT) with the Fast Gradient Sign Method (FGSM) where the test robustness steeply declines over just one epoch in the training stage. Prior work has attributed this loss in robustness to a sharp decrease in $\textit{local linearity}$ of the neural network with respect to the input space, and has demonstrated that introducing a local linearity measure as a regularization term prevents catastrophic overfitting. Using a simple neural network architecture, we experimentally demonstrate that maintaining high local linearity might be $\textit{sufficient}$ to prevent catastrophic overfitting but is not $\textit{necessary.}$ Further, inspired by Parseval networks, we introduce a regularization term to AT with FGSM to make the weight matrices of the network orthogonal and study the connection between orthogonality of the network weights and local linearity. Lastly, we identify the $\textit{double descent}$ phenomenon during the adversarial training process.

cs.LG↗