sequence
6845 papers tagged with this keyword
Better Neural Network Expressivity: Subdividing the Simplex
This work studies the expressivity of ReLU neural networks with a focus on their depth. A sequence of previous works showed that $\lceil \log_2(n+1) \rceil$ hidden layers are sufficient to compute all continuous piecewise linear (CPWL) functions on $\mathbb{R}^n$. Hertrich, Basu, Di Summa, and Skutella (NeurIPS'21 / SIDMA'23) conjectured that this result is optimal in the sense that there are CPWL functions on $\mathbb{R}^n$, like the maximum function, that require this depth. We disprove the conjecture and show that $\lceil\log_3(n-1)\rceil+1$ hidden layers are sufficient to compute all CPWL functions on $\mathbb{R}^n$.
A key step in the proof is that ReLU neural networks with two hidden layers can exactly represent the maximum function of five inputs. More generally, we show that $\lceil\log_3(n-2)\rceil+1$ hidden layers are sufficient to compute the maximum of $n\geq 4$ numbers. Our constructions almost match the $\lceil\log_3(n)\rceil$ lower bound of Averkov, Hojny, and Merkert (ICLR'25) in the special case of ReLU networks with weights that are decimal fractions. The constructions have a geometric interpretation via polyhedral subdivisions of the simplex into ``easier'' polytopes.
Connections Between Frames with Rational Eigensteps and Semistandard Young Tableaux
In this paper, we explore a correspondence between frames with rational eigensteps and semistandard Young tableaux (SSYT), via the relation assigning a Gelfand-Tsetlin pattern to a frame via the frame's eigensteps. We will identify how certain key structures in SSYTs correlate with particular frame properties. For example, the weight of an SSYT yields the sequence of norms of any compatible frame. Additionally, this correspondence leads to a novel way to construct the eigensteps of a frame coming solely from tableaux. This is an alternative to the Top Kill algorithm which may be viewed as a combinatorial reinterpretation of the algorithm. We further employ other combinatorial techniques such as the boxcomp method to generate a ``complement" SSYT. On the frame side, this corresponds to a tight frame's Naimark complement as well as to a generalization of the Naimark complement for non-tight frames. Further research points to an analysis of equiangular tight frames and their corresponding tableaux, as well as using more combinatorial operations to further analyze frames.
Independent Set Enumeration in King Graphs by Tensor Network Contractions
This paper discusses the enumeration of independent sets in king graphs of size $m \times n$, based on the tensor network contractions algorithm given in reference~\cite{tilEnum}. We transform the problem into Wang tiling enumeration within an $(m+1) \times (n+1)$ rectangle and compute the results for all cases where $m + n \leq 79$ using tensor network contraction algorithm, and provided an approximation for larger $m, n$.
Using the same algorithm, we also enumerated independent sets with vertex number restrictions. Based on the results, we analyzed the vertex number that maximize the enumeration for each pair $(m, n)$. Additionally, we compute the corresponding weighted enumeration, where each independent set is weighted by the number of its vertices (i.e., the total sum of vertices over all independent sets). The approximations for larger $m, n$ are given as well.
Our results have added thousands of new items to the OEIS sequences A089980 and A193580. In addition, the combinatorial problems above are closely related to the hard-core model in physics. We estimate some important constants based on the existing results, and the relative error between our estimation of the entropy constant and the existing results is less than $10^{-9}$.
Bounds on Trees with Topological Indices Among Degree Sequence
In this paper, we investigate The relationship between the Albertson index and the first Zagreb index for trees. For a tree $T=(V,E)$ with $n=|V|$ vertices and $m=|E|$ edges, we provide several bounds and exact formulas for these two topological indices, and we show that the Albertson index $\irr(T)$ and the first Zagreb index $M_1(T)$ satisfy the association \[ \operatorname{irr}(T)=d_1^2+d_n^2+(n-2)\left(\frac{Δ+ δ}{2}\right)^2+\sum_{i=2}^{n-1} d_i+d_n - d_1-2n-2.\] Our goal of this paper is provide a topological indices, Albertson index, Sigma index among a degree sequence $\mathscr{D}=(d_1,\dots,d_n)$ where it is non-increasing and non-decreasing of tree $T$.
A local framework for proving combinatorial matrix inversion theorems
Combinatorial transition matrices arise frequently in the theory of symmetric functions and their generalizations. The entries of such matrices often count signed, weighted combinatorial structures such as semistandard tableaux, rim-hook tableaux, or brick tabloids. Bijective proofs that two such matrices are inverses of each other may be difficult to find. This paper presents a general framework for proving such inversion results in the case where the combinatorial objects are built up recursively by successively adding some incremental structure such as a single horizontal strip or rim-hook. In this setting, we show that a sequence of matrix inversion results $A_nB_n=I$ can be reduced to a certain ``local'' identity involving the incremental structures. Here, $A_n$ and $B_n$ are matrices that might be non-square, and the columns of $A_n$ and the rows of $B_n$ indexed by compositions of $n$. We illustrate the general theory with four classical applications involving the Kostka matrices, the character tables of the symmetric group, incidence matrices for composition posets, and matrices counting brick tabloids. We obtain a new, canonical bijective proof of an inversion result for rectangular Kostka matrices, which complements the proof for the square case due to Eğecioğlu and Remmel. We also give a new bijective proof of the orthogonality result for the irreducible $S_n$-characters that is shorter than the original version due to White.
A note on iterated sumsets races
This short note answers a question raised by Nathanson \cite{Nath25} about "races" between iterated sumsets. We prove that for any integer $n$, there are finite sets of integers $A$ and $B$ with same diameter such that the signs of the elements of the sequence $(|hA|-|hB|)_h$ changes at least $n$ times. Kravitz proved in \cite{Kravitz} a much better result. This brief and modest note may serve as a stepping stone towards his work.
Generalized break divisors and triangulations of Lawrence polytopes
Let $G$ be a connected graph of genus $g$. The Picard group of degree $g$, $\text{Pic}^g(G)$, is the set of equivalence classes of divisors on $G$ of degree $g$, where two divisors are equivalent if one can be reached from the other through a sequence of chip-firing moves. We construct sets of representatives of the equivalence classes in $\text{Pic}^g(G)$ by defining a function $I_G$ on the spanning trees of $G$ from a triangulation of the Lawrence polytope of the cographic matroid $\mathcal{M}^\ast(G)$. Additionally, such sets of representatives correspond to stability conditions on the nodal curve dual to the graph $G$. We show that $I_G$ that are constructed from regular triangulations of Lawrence polytope correspond to classical stability conditions, which are induced by generic real-valued divisors on $G$.
Small hitting sets for longest paths and cycles
Motivated by an old question of Gallai (1966) on the intersection of longest paths in a graph and the well-known conjectures of Lovász (1969) and Thomassen (1978) on the maximum length of paths and cycles in vertex-transitive graphs, we present improved bounds for the parameters $\mathrm{lpt}(G)$ and $\mathrm{lct}(G)$, defined as the minimum size of a set of vertices in a graph $G$ hitting all longest paths (cycles, respectively). First, we show that every connected graph $G$ on $n$ vertices satisfies $\mathrm{lpt}(G)\le \sqrt{8n}$, and $\mathrm{lct}(G)\le \sqrt{8n}$ if $G$ is additionally $2$-connected. This improves a sequence of earlier bounds for these problems, with the previous state of the art being $O(n^{2/3})$. Second, we show that every connected graph $G$ satisfies $\mathrm{lpt}(G)\le O(\ell^{5/9})$, where $\ell$ denotes the maximum length of a path in $G$. As an immediate application of this latter bound, we present further progress towards Lovász' and Thomassen's conjectures: We show that every connected vertex-transitive graph of order $n$ contains a cycle (and path) of length $Ω(n^{9/14})$. This improves the previous best bound of the form $Ω(n^{13/21})$. Interestingly, our proofs make use of several concepts and results from structural graph theory, such as a result of Robertson and Seymour (1990) on transactions in societies and Tutte's $2$-separator theorem.
Topological Indices Among Strong Support Vertex
In this paper, we provide the irregularity properties of trees with strong support vertex by analyzing two prominent topological indices: the Albertson index and the Sigma index. We further establish extremal bounds for both indices across families of trees defined by given degree sequences. Let $T_1$ and $T_2$ be a star trees of order $n$, where $T_1 \cong T_2$, then, we provide Albertson index of $T_1 \cong T_2$. Let $\mathcal{T}_{n, Δ}$ be a class of trees with $n$ vertices, there are a tree $T^{\prime} \in \mathcal{T}_{n, Δ}$ such that $\irr(T^{\prime}) < \irr(T)$.
Reconfiguration of List Colourings
Given a proper (list) colouring of a graph $G$, a recolouring step changes the colour at a single vertex to another colour (in its list) that is currently unused on its neighbours, hence maintaining a proper colouring. Suppose that each vertex $v$ has its own private list $L(v)$ of allowed colours such that $|L(v)|\ge \mbox{deg}(v)+1$. We prove that if $G$ is connected and its maximum degree $Δ$ is at least $3$, then for any two proper $L$-colourings in which at least one vertex can be recoloured, one can be transformed to the other by a sequence of $O(|V(G)|^2)$ recolouring steps. We also show that reducing the list-size of a single vertex $w$ to $\mbox{deg}(w)$ can lead to situations where the space of proper $L$-colourings is `shattered'. Our results can be interpreted as showing a sharp phase transition in the Glauber dynamics of proper $L$-colourings of graphs. This constitutes a `local' strengthening and generalisation of a result of Feghali, Johnson, and Paulusma, which considered the situation where the lists are all identical to $\{1,\ldots,Δ+1\}$.
When are Hopf algebras determined by integer sequences?
We study the category of graded Hopf algebras that are free noncommutative, cocommutative, graded and connected from the perspective of the sequences of dimensions of the graded pieces. We show that a Hopf algebra exists with a given sequence of graded dimensions if and only if the ``INVERTi'' transformation of the sequence is nonnegative. We give conditions on the sequences of graded dimensions for two Hopf algebras $H$ and $K$ in this category under which there exists a surjective homomorphism from $H$ to $K$. We also give conditions such that an isomorphic copy of $H$ occurs as a Hopf subalgebra of $K$.
Positivity in the shadow of Hodge index theorem
Taking a compact Kähler manifold as playground, we explore the powerfulness of Hodge index theorem. A main object is the Lorentzian classes on a compact Kähler manifold, behind which the characterization via Lorentzian polynomials over the Kähler cone and hence the validity of Hodge index theorem. Along the exploration, we discover several applications in complex geometry that may be unexpected before. (1) For a Lefschetz type operator given by the complete intersection of nef classes, we give a complete characterization of its kernel face against the pseudo-effective cone. (2) We provide a new approach to Teissier's proportionality problem from the validity of hard Lefschetz property. This perspective enables us to establish the extremals for the Brunn-Minkowski inequality on a strictly Lorentzian class, and thus also characterize the most extremal case for a log-concavity sequence given by the intersection numbers of two nef classes. These Lorentzian classes include the fundamental classes of smooth projective varieties or compact Kähler manifolds as typical examples, hence our result extends Boucksom-Favre-Jonsson's and Fu-Xiao's results in respective settings to broader contexts, e.g. certain algebraic cycle classes given by reducible subvarieties. (3) Furthermore, we also strengthen the proportionality characterization by comparing various quantitative deficits and establishing stability estimates. Two quantitative sharper stability estimates with close relation with complex Monge--Ampère equations and Newton-Okounkov bodies are also discussed.
Counting subgraphs of coloring graphs using shadow graphs
Given a graph $G$, the $k$-coloring graph $\mathcal{C}_k(G)$ is constructed by selecting proper $k$-colorings of $G$ as vertices, with an edge between two colorings if they differ in the color of exactly one vertex. The number of vertices in $\mathcal{C}_k(G)$ is the famous chromatic polynomial of $G$. Asgarli, Krehbiel, Levinson and Russell showed that for any subgraph $H$, the number of induced copies of $H$ in $\mathcal{C}_k(G)$ is a polynomial function in $k$. Hogan, Scott, Tamitegama, and Tan found a shorter proof for polynomiality of these chromatic $H$-polynomials. In this paper, we provide a method of constructing these polynomials explicitly in terms of chromatic polynomials of shadow graphs. We illustrate the practicality of our formulas by computing an explicit formula for $H$-polynomial for trees when $H=Q_d$ is an arbitrary hypercube, a task which does not seem approachable from previous methods. The coefficients of the resulting polynomials feature generalized degree sequences introduced by Crew. In the special case when $H=P_2$, the corresponding polynomial is dubbed the chromatic pairs polynomial. We present a pair of graphs $G_1$ and $G_2$ sharing the same chromatic pairs polynomial but different chromatic polynomials, disproving a conjecture raised by Asgarli, Krehbiel, Levinson and Russell.
Analytic properties arising from the Baxter numbers
Baxter numbers are known as the enumeration of Baxter permutations and numerous other discrete structures, playing a significant role across combinatorics, algebra, and analysis. In this paper, we focus on the analytic properties related to Baxter numbers. We prove that the descent polynomials of Baxter permutations have interlacing zeros, which is a property stronger than real-rootedness. Our approach is based on Dilks' framework of $(q,t)$-Hoggatt sums, which is a $q$-analog for Baxter permutations. Within this framework, we show that the family of $(1,t)$-Hoggatt sums satisfies the interlacing property using fundamental results on Hadamard products of polynomials. For Baxter numbers, we prove their asymptotic $r$-log-convexity via asymptotic expansions of $P$-recursive sequences. In particular, we confirm their $2$-log-convexity using symbolic computation techniques.
Scheduled Jacobian Chaining
This paper addresses the efficient computation of Jacobian matrices for programs composed of sequential differentiable subprograms. By representing the overall Jacobian as a chain product of the Jacobians of these subprograms, we reduce the problem to optimizing the sequence of matrix multiplications, known as the Jacobian Matrix Chain Product problem. Solutions to this problem yield "optimal bracketings", which induce a precedence-constraint scheduling problem. We investigate the inherent parallelism in the solutions and develop a new dynamic programming algorithm as a heuristic that incorporates the scheduling. To assess its performance, we benchmark it against the global optimum, which is computed via a branch-and-bound algorithm.
A note on the irrationality of $ζ_2(5)$
In a spirit of Apéry's proof of the irrationality of $ζ(3)$, we construct a sequence $p_n/q_n$ of rational approximations to the $2$-adic zeta value $ζ_2(5)$ which satisfy $0 < |ζ_2(5)-p_n/q_n|_2 < \max\{|p_n|,|q_n|\}^{-1-δ}$ for an explicit constant $δ>0$. This leads to a new proof of the irrationality of $ζ_2(5)$, the result established recently by Calegari, Dimitrov and Tang using a different method. Furthermore, our approximations allow us to obtain an upper bound for the irrationality measure of this $2$-adic quantity; namely, we show that $μ(ζ_2(5)) \le (16\log2)/(8\log2-5) = 20.342\dots$.
Pattern avoidance in revised ascent sequences
Inspired by the definition of modified ascent sequences, we introduce a new class of integer sequences called revised ascent sequences. These sequences are defined as Cayley permutations where each entry is a leftmost occurrence if and only if it serves as an ascent bottom. We construct a bijection between ascent sequences and revised ascent sequences by adapting the classic hat map, which transforms ascent sequences into modified ascent sequences. Additionally, we investigate revised ascent sequences that avoid a single pattern, leading to a wealth of enumerative results. Our main techniques include the use of bijections, generating trees, generating functions, and the kernel method.
The geometry of zonotopal algebras II: Orlik--Terao algebras and Schubert varieties
Zonotopal algebras, introduced by Postnikov--Shapiro--Shapiro, Ardila--Postnikov, and Holtz--Ron, show up in many different contexts, including approximation theory, representation theory, Donaldson--Thomas theory, and hypertoric geometry. In the first half of this paper, we construct a perfect pairing between the internal zonotopal algebra of a linear space and the reduced Orlik--Terao algebra of the Gale dual linear space. As an application, we prove a conjecture of Moseley--Proudfoot--Young that relates the reduced Orlik--Terao algebra of a graph to the cohomology of a certain configuration space. In the second half of the paper, we interpret the Macaulay inverse system of a zonotopal algebra as the space of sections of a sheaf on the Schubert variety of a linear space. As an application of this, we use an equivariant resolution of the structure sheaf of the Schubert variety inside of a product of projective lines to produce an exact sequence relating internal and external zonotopal algebras.
On increasing sequences formed by points from a random finite subset of a hypercube
Consider $S$, a set of $n$ points chosen uniformly at random and independently from the unit hypercube of dimension $t>2$. Order $S$ by using the Cartesian product of the $t$ standard orders of $[0,1]$. We determine a constant $\bar x(t)<e$ such that, with probability $\ge 1-\exp(-Θ(\eps)n^{1/t})$, cardinality of a largest subset of comparable points is at most $(\bar x(t)+\eps)n^{1/t}$. The bound $\bar x(t)$ complements an explicit lower bound obtained by Bollobás and Winkler in 1982. Furthermore, we use Dilworth's theorem on partitions of a set into chains to prove that the cardinality of a largest antichain, i. e. a largest subset of incomparable points, is at least $(1-\eps) (n/e)^{1-1/t}$ with probability exponentially close to $1$.
A note on the number of distinct elements and zero-sum subsequence lengths in cyclic groups
In this short note we investigate zero-sum sequences in finite abelian groups, examining the relationship between the sequence's support size, that is the number of distinct elements, and its properties concerning zero-sums. In particular, for sequences $S$ in a cyclic group, we establish a direct connection between $MZ(S)$, the length of the shortest nonempty subsequence summing to zero and the number of distinct values in $S$. Our results reveal that sequences with larger support must contain shorter non-empty zero-sum subsequences, in line with classical zero-sum results. Additionally, we present one application of our main result to a factorization of ideals problem in rings of integers of a number field.