-
Improved Bounds for Point Selections and Halving Hyperplanes in Higher Dimensions
Authors:
Natan Rubin
Abstract:
Let $(P,E)$ be a $(d+1)$-uniform geometric hypergraph, where $P$ is an $n$-point set in general position in $\mathbb{R}^d$ and $E\subseteq {P\choose d+1}$ is a collection of $ε{n\choose d+1}$ $d$-dimensional simplices with vertices in $P$, for $0<ε\leq 1$. We show that there is a point $x\in {\mathbb R}^d$ that pierces $\displaystyle Ω\left(ε^{(d^4+d)(d+1)+δ}{n\choose d+1}\right)$ simplices in…
▽ More
Let $(P,E)$ be a $(d+1)$-uniform geometric hypergraph, where $P$ is an $n$-point set in general position in $\mathbb{R}^d$ and $E\subseteq {P\choose d+1}$ is a collection of $ε{n\choose d+1}$ $d$-dimensional simplices with vertices in $P$, for $0<ε\leq 1$. We show that there is a point $x\in {\mathbb R}^d$ that pierces $\displaystyle Ω\left(ε^{(d^4+d)(d+1)+δ}{n\choose d+1}\right)$ simplices in $E$, for any fixed $δ>0$. This is a dramatic improvement in all dimensions $d\geq 3$, over the previous lower bounds of the general form $\displaystyle ε^{(cd)^{d+1}}n^{d+1}$, which date back to the seminal 1991 work of Alon, Bárány, Füredi and Kleitman.
As a result, any $n$-point set in general position in $\mathbb{R}^d$ admits only $\displaystyle O\left(n^{d-\frac{1}{d(d-1)^4+d(d-1)}+δ}\right)$ halving hyperplanes, for any $δ>0$, which is a significant improvement over the previously best known bound $\displaystyle O\left(n^{d-\frac{1}{(2d)^{d}}}\right)$ in all dimensions $d\geq 5$.
An essential ingredient of our proof is the following semi-algebraic Turán-type result of independent interest: Let $(V_1,\ldots,V_k,E)$ be a hypergraph of bounded semi-algebraic description complexity in ${\mathbb R}^d$ that satisfies $|E|\geq \varepsilon |V_1|\cdot\ldots \cdot |V_k|$ for some $\varepsilon>0$. Then there exist subsets $W_i\subseteq V_i$ that satisfy $W_1\times W_2\times\ldots\times W_k\subseteq E$, and $|W_1|\cdot\ldots\cdots|W_k|=Ω\left(\varepsilon^{d(k-1)+1}|V_1|\cdot |V_2|\cdot\ldots\cdot|V_k|\right)$.
△ Less
Submitted 1 March, 2024;
originally announced March 2024.
-
AXI4MLIR: User-Driven Automatic Host Code Generation for Custom AXI-Based Accelerators
Authors:
Nicolas Bohm Agostini,
Jude Haris,
Perry Gibson,
Malith Jayaweera,
Norm Rubin,
Antonino Tumeo,
José L. Abellán,
José Cano,
David Kaeli
Abstract:
This paper addresses the need for automatic and efficient generation of host driver code for arbitrary custom AXI-based accelerators targeting linear algebra algorithms, an important workload in various applications, including machine learning and scientific computing. While existing tools have focused on automating accelerator prototy**, little attention has been paid to the host-accelerator in…
▽ More
This paper addresses the need for automatic and efficient generation of host driver code for arbitrary custom AXI-based accelerators targeting linear algebra algorithms, an important workload in various applications, including machine learning and scientific computing. While existing tools have focused on automating accelerator prototy**, little attention has been paid to the host-accelerator interaction. This paper introduces AXI4MLIR, an extension of the MLIR compiler framework designed to facilitate the automated generation of host-accelerator driver code. With new MLIR attributes and transformations, AXI4MLIR empowers users to specify accelerator features (including their instructions) and communication patterns and exploit the host memory hierarchy. We demonstrate AXI4MLIR's versatility across different types of accelerators and problems, showcasing significant CPU cache reference reductions (up to 56%) and up to a 1.65x speedup compared to manually optimized driver code implementations. AXI4MLIR implementation is open-source and available at: https://github.com/AXI4MLIR/axi4mlir.
△ Less
Submitted 22 December, 2023;
originally announced December 2023.
-
Grokking as a First Order Phase Transition in Two Layer Networks
Authors:
Noa Rubin,
Inbar Seroussi,
Zohar Ringel
Abstract:
A key property of deep neural networks (DNNs) is their ability to learn new features during training. This intriguing aspect of deep learning stands out most clearly in recently reported Grokking phenomena. While mainly reflected as a sudden increase in test accuracy, Grokking is also believed to be a beyond lazy-learning/Gaussian Process (GP) phenomenon involving feature learning. Here we apply a…
▽ More
A key property of deep neural networks (DNNs) is their ability to learn new features during training. This intriguing aspect of deep learning stands out most clearly in recently reported Grokking phenomena. While mainly reflected as a sudden increase in test accuracy, Grokking is also believed to be a beyond lazy-learning/Gaussian Process (GP) phenomenon involving feature learning. Here we apply a recent development in the theory of feature learning, the adaptive kernel approach, to two teacher-student models with cubic-polynomial and modular addition teachers. We provide analytical predictions on feature learning and Grokking properties of these models and demonstrate a map** between Grokking and the theory of phase transitions. We show that after Grokking, the state of the DNN is analogous to the mixed phase following a first-order phase transition. In this mixed phase, the DNN generates useful internal representations of the teacher that are sharply distinct from those before the transition.
△ Less
Submitted 5 May, 2024; v1 submitted 5 October, 2023;
originally announced October 2023.
-
Expanding the reach of quantum optimization with fermionic embeddings
Authors:
Andrew Zhao,
Nicholas C. Rubin
Abstract:
Quadratic programming over orthogonal matrices encompasses a broad class of hard optimization problems that do not have an efficient quantum representation. Such problems are instances of the little noncommutative Grothendieck problem (LNCG), a generalization of binary quadratic programs to continuous, noncommutative variables. In this work, we establish a natural embedding for this class of LNCG…
▽ More
Quadratic programming over orthogonal matrices encompasses a broad class of hard optimization problems that do not have an efficient quantum representation. Such problems are instances of the little noncommutative Grothendieck problem (LNCG), a generalization of binary quadratic programs to continuous, noncommutative variables. In this work, we establish a natural embedding for this class of LNCG problems onto a fermionic Hamiltonian, thereby enabling the study of this classical problem with the tools of quantum information. This embedding is accomplished by identifying the orthogonal group with its double cover, which can be represented as fermionic quantum states. Correspondingly, the embedded LNCG Hamiltonian is a two-body fermion model. Determining extremal states of this Hamiltonian provides an outer approximation to the original problem, a quantum analogue to classical semidefinite relaxations. In particular, when optimizing over the special orthogonal group our quantum relaxation obeys additional, powerful constraints based on the convex hull of rotation matrices. The classical size of this convex-hull representation is exponential in matrix dimension, whereas our quantum representation requires only a linear number of qubits. Finally, to project the relaxed solution back into the feasible space, we propose rounding procedures which return orthogonal matrices from appropriate measurements of the quantum state. Through numerical experiments we provide evidence that this rounded quantum relaxation can produce high-quality approximations.
△ Less
Submitted 9 June, 2023; v1 submitted 4 January, 2023;
originally announced January 2023.
-
Generating GPU Compiler Heuristics using Reinforcement Learning
Authors:
Ian Colbert,
Jake Daly,
Norm Rubin
Abstract:
GPU compilers are complex software programs with many optimizations specific to target hardware. These optimizations are often controlled by heuristics hand-designed by compiler experts using time- and resource-intensive processes. In this paper, we developed a GPU compiler autotuning framework that uses off-policy deep reinforcement learning to generate heuristics that improve the frame rates of…
▽ More
GPU compilers are complex software programs with many optimizations specific to target hardware. These optimizations are often controlled by heuristics hand-designed by compiler experts using time- and resource-intensive processes. In this paper, we developed a GPU compiler autotuning framework that uses off-policy deep reinforcement learning to generate heuristics that improve the frame rates of graphics applications. Furthermore, we demonstrate the resilience of these learned heuristics to frequent compiler updates by analyzing their stability across a year of code check-ins without retraining. We show that our machine learning-based compiler autotuning framework matches or surpasses the frame rates for 98% of graphics benchmarks with an average uplift of 1.6% up to 15.8%.
△ Less
Submitted 23 November, 2021;
originally announced November 2021.
-
Stronger Bounds for Weak Epsilon-Nets in Higher Dimensions
Authors:
Natan Rubin
Abstract:
Given a finite point set $P$ in ${\mathbb R}^d$, and $ε>0$ we say that $N\subseteq{ \mathbb R}^d$ is a weak $ε$-net if it pierces every convex set $K$ with $|K\cap P|\geq ε|P|$. We show that for any finite point set in dimension $d\geq 3$, and any $ε>0$, one can construct a weak $ε$-net whose cardinality is $\displaystyle O^*\left(\frac{1}{ε^{2.558}}\right)$ in dimension $d=3$, and…
▽ More
Given a finite point set $P$ in ${\mathbb R}^d$, and $ε>0$ we say that $N\subseteq{ \mathbb R}^d$ is a weak $ε$-net if it pierces every convex set $K$ with $|K\cap P|\geq ε|P|$. We show that for any finite point set in dimension $d\geq 3$, and any $ε>0$, one can construct a weak $ε$-net whose cardinality is $\displaystyle O^*\left(\frac{1}{ε^{2.558}}\right)$ in dimension $d=3$, and $\displaystyle o\left(\frac{1}{ε^{d-1/2}}\right)$ in all dimensions $d\geq 4$.
To be precise, our weak $ε$-net has cardinality $\displaystyle O\left(\frac{1}{ε^{α_d+γ}}\right)$ for any $γ>0$, with
$$
α_d=
\left\{
\begin{array}{l}
2.558 & \text{if} \ d=3
\\3.48 & \text{if} \ d=4
\\\left(d+\sqrt{d^2-2d}\right)/2 & \text{if} \ d\geq 5.
\end{array}\right\}
$$
This is the first significant improvement of the bound of $\displaystyle \tilde{O}\left(\frac{1}{ε^d}\right)$ that was obtained in 1993 by Chazelle, Edelsbrunner, Grigni, Guibas, Sharir, and Welzl for general point sets in dimension $d\geq 3$.
△ Less
Submitted 26 December, 2023; v1 submitted 26 April, 2021;
originally announced April 2021.
-
Integer and Constraint Programming Revisited for Mutually Orthogonal Latin Squares
Authors:
Noah Rubin,
Curtis Bright,
Kevin K. H. Cheung,
Brett Stevens
Abstract:
In this paper we provide results on using integer programming (IP) and constraint programming (CP) to search for sets of mutually orthogonal latin squares (MOLS). Both programming paradigms have previously successfully been used to search for MOLS, but solvers for IP and CP solvers have significantly improved in recent years and data on how modern IP and CP solvers perform on the MOLS problem is l…
▽ More
In this paper we provide results on using integer programming (IP) and constraint programming (CP) to search for sets of mutually orthogonal latin squares (MOLS). Both programming paradigms have previously successfully been used to search for MOLS, but solvers for IP and CP solvers have significantly improved in recent years and data on how modern IP and CP solvers perform on the MOLS problem is lacking. Using state-of-the-art solvers as black boxes we were able to quickly find pairs of MOLS (or prove their nonexistence) in all orders up to ten. Moreover, we improve the effectiveness of the solvers by formulating an extended symmetry breaking method as well as an improvement to the straightforward CP encoding. We also analyze the effectiveness of using CP and IP solvers to search for triples of MOLS, compare our timings to those which have been previously published, and estimate the running time of using this approach to resolve the longstanding open problem of determining the existence of a triple of MOLS of order ten.
△ Less
Submitted 19 March, 2021;
originally announced March 2021.
-
Planar Point Sets Determine Many Pairwise Crossing Segments
Authors:
János Pach,
Natan Rubin,
Gábor Tardos
Abstract:
We show that any set of $n$ points in general position in the plane determines $n^{1-o(1)}$ pairwise crossing segments. The best previously known lower bound, $Ω\left(\sqrt n\right)$, was proved more than 25 years ago by Aronov, Erd\H os, Goddard, Kleitman, Klugerman, Pach, and Schulman. Our proof is fully constructive, and extends to dense geometric graphs.
We show that any set of $n$ points in general position in the plane determines $n^{1-o(1)}$ pairwise crossing segments. The best previously known lower bound, $Ω\left(\sqrt n\right)$, was proved more than 25 years ago by Aronov, Erd\H os, Goddard, Kleitman, Klugerman, Pach, and Schulman. Our proof is fully constructive, and extends to dense geometric graphs.
△ Less
Submitted 30 April, 2023; v1 submitted 18 April, 2019;
originally announced April 2019.
-
An Improved Bound for Weak Epsilon-Nets in the Plane
Authors:
Natan Rubin
Abstract:
We show that for any finite set $P$ of points in the plane and $ε>0$ there exist $\displaystyle O\left(\frac{1}{ε^{3/2+γ}}\right)$ points in ${\mathbb{R}}^2$, for arbitrary small $γ>0$, that pierce every convex set $K$ with $|K\cap P|\geq ε|P|$. This is the first improvement of the bound of $\displaystyle O\left(\frac{1}{ε^2}\right)$ that was obtained in 1992 by Alon, Bárány, Füredi and Kleitman f…
▽ More
We show that for any finite set $P$ of points in the plane and $ε>0$ there exist $\displaystyle O\left(\frac{1}{ε^{3/2+γ}}\right)$ points in ${\mathbb{R}}^2$, for arbitrary small $γ>0$, that pierce every convex set $K$ with $|K\cap P|\geq ε|P|$. This is the first improvement of the bound of $\displaystyle O\left(\frac{1}{ε^2}\right)$ that was obtained in 1992 by Alon, Bárány, Füredi and Kleitman for general point sets in the plane.
△ Less
Submitted 21 July, 2022; v1 submitted 8 August, 2018;
originally announced August 2018.
-
Further Consequences of the Colorful Helly Hypothesis
Authors:
Leonardo Martínez-Sandoval,
Edgardo Roldán-Pensado,
Natan Rubin
Abstract:
Let $\mathcal{F}$ be a family of convex sets in ${\mathbb R}^d$, which are colored with $d+1$ colors. We say that $\mathcal{F}$ satisfies the Colorful Helly Property if every rainbow selection of $d+1$ sets, one set from each color class, has a non-empty common intersection. The Colorful Helly Theorem of Lovász states that for any such colorful family $\mathcal{F}$ there is a color class…
▽ More
Let $\mathcal{F}$ be a family of convex sets in ${\mathbb R}^d$, which are colored with $d+1$ colors. We say that $\mathcal{F}$ satisfies the Colorful Helly Property if every rainbow selection of $d+1$ sets, one set from each color class, has a non-empty common intersection. The Colorful Helly Theorem of Lovász states that for any such colorful family $\mathcal{F}$ there is a color class $\mathcal{F}_i\subset \mathcal{F}$, for $1\leq i\leq d+1$, whose sets have a non-empty intersection.
We establish further consequences of the Colorful Helly hypothesis. In particular, we show that for each dimension $d\geq 2$ there exist numbers $f(d)$ and $g(d)$ with the following property: either one can find an additional color class whose sets can be pierced by $f(d)$ points, or all the sets in $\mathcal{F}$ can be crossed by $g(d)$ lines.
△ Less
Submitted 16 March, 2018;
originally announced March 2018.
-
A Crossing Lemma for Jordan Curves
Authors:
János Pach,
Natan Rubin,
Gábor Tardos
Abstract:
If two Jordan curves in the plane have precisely one point in common, and there they do not properly cross, then the common point is called a {\em touching point}. The main result of this paper is a Crossing Lemma for simple curves: Let $X$ and $T$ stand for the sets of intersection points and touching points, respectively, in a family of $n$ simple curves in the plane, no three of which pass thro…
▽ More
If two Jordan curves in the plane have precisely one point in common, and there they do not properly cross, then the common point is called a {\em touching point}. The main result of this paper is a Crossing Lemma for simple curves: Let $X$ and $T$ stand for the sets of intersection points and touching points, respectively, in a family of $n$ simple curves in the plane, no three of which pass through the same point. If $|T|>cn$, for some fixed constant $c>0$, then we prove that $|X|=Ω(|T|(\log\log(|T|/n))^{1/504})$. In particular, if $|T|/n\rightarrow\infty$, then the number of intersection points is much larger than the number of touching points.
As a corollary, we confirm the following long-standing conjecture of Richter and Thomassen: The total number of intersection points between $n$ pairwise intersecting simple closed (i.e., Jordan) curves in the plane, no three of which pass through the same point, is at least $(1-o(1))n^2$.
△ Less
Submitted 7 August, 2017;
originally announced August 2017.
-
Beyond the Richter-Thomassen Conjecture
Authors:
János Pach,
Natan Rubin,
Gábor Tardos
Abstract:
If two closed Jordan curves in the plane have precisely one point in common, then it is called a {\em touching point}. All other intersection points are called {\em crossing points}. The main result of this paper is a Crossing Lemma for closed curves: In any family of $n$ pairwise intersecting simple closed curves in the plane, no three of which pass through the same point, the number of crossing…
▽ More
If two closed Jordan curves in the plane have precisely one point in common, then it is called a {\em touching point}. All other intersection points are called {\em crossing points}. The main result of this paper is a Crossing Lemma for closed curves: In any family of $n$ pairwise intersecting simple closed curves in the plane, no three of which pass through the same point, the number of crossing points exceeds the number of touching points by a factor of at least $Ω((\log\log n)^{1/8})$.
As a corollary, we prove the following long-standing conjecture of Richter and Thomassen: The total number of intersection points between any $n$ pairwise intersecting simple closed curves in the plane, no three of which pass through the same point, is at least $(1-o(1))n^2$.
△ Less
Submitted 7 July, 2015; v1 submitted 30 April, 2015;
originally announced April 2015.
-
Stable Delaunay Graphs
Authors:
Pankaj K. Agarwal,
Jie Gao,
Leonidas J. Guibas,
Haim Kaplan,
Natan Rubin,
Micha Sharir
Abstract:
Let $P$ be a set of $n$ points in $\mathrm{R}^2$, and let $\mathrm{DT}(P)$ denote its Euclidean Delaunay triangulation. We introduce the notion of an edge of $\mathrm{DT}(P)$ being {\it stable}. Defined in terms of a parameter $α>0$, a Delaunay edge $pq$ is called $α$-stable, if the (equal) angles at which $p$ and $q$ see the corresponding Voronoi edge $e_{pq}$ are at least $α$. A subgraph $G$ of…
▽ More
Let $P$ be a set of $n$ points in $\mathrm{R}^2$, and let $\mathrm{DT}(P)$ denote its Euclidean Delaunay triangulation. We introduce the notion of an edge of $\mathrm{DT}(P)$ being {\it stable}. Defined in terms of a parameter $α>0$, a Delaunay edge $pq$ is called $α$-stable, if the (equal) angles at which $p$ and $q$ see the corresponding Voronoi edge $e_{pq}$ are at least $α$. A subgraph $G$ of $\mathrm{DT}(P)$ is called {\it $(cα, α)$-stable Delaunay graph} ($\mathrm{SDG}$ in short), for some constant $c \ge 1$, if every edge in $G$ is $α$-stable and every $cα$-stable of $\mathrm{DT}(P)$ is in $G$.
We show that if an edge is stable in the Euclidean Delaunay triangulation of $P$, then it is also a stable edge, though for a different value of $α$, in the Delaunay triangulation of $P$ under any convex distance function that is sufficiently close to the Euclidean norm, and vice-versa. In particular, a $6α$-stable edge in $\mathrm{DT}(P)$ is $α$-stable in the Delaunay triangulation under the distance function induced by a regular $k$-gon for $k \ge 2π/α$, and vice-versa. Exploiting this relationship and the analysis in~\cite{polydel}, we present a linear-size kinetic data structure (KDS) for maintaining an $(8α,α)$-$\mathrm{SDG}$ as the points of $P$ move. If the points move along algebraic trajectories of bounded degree, the KDS processes nearly quadratic events during the motion, each of which can processed in $O(\log n)$ time. Finally, we show that a number of useful properties of $\mathrm{DT}(P)$ are retained by $\mathrm{SDG}$ of $P$.
△ Less
Submitted 26 April, 2015;
originally announced April 2015.
-
On the Richter-Thomassen Conjecture about Pairwise Intersecting Closed Curves
Authors:
János Pach,
Natan Rubin,
Gábor Tardos
Abstract:
A long standing conjecture of Richter and Thomassen states that the total number of intersection points between any $n$ simple closed Jordan curves in the plane, so that any pair of them intersect and no three curves pass through the same point, is at least $(1-o(1))n^2$.
We confirm the above conjecture in several important cases, including the case (1) when all curves are convex, and (2) when t…
▽ More
A long standing conjecture of Richter and Thomassen states that the total number of intersection points between any $n$ simple closed Jordan curves in the plane, so that any pair of them intersect and no three curves pass through the same point, is at least $(1-o(1))n^2$.
We confirm the above conjecture in several important cases, including the case (1) when all curves are convex, and (2) when the family of curves can be partitioned into two equal classes such that each curve from the first class is touching every curve from the second class. (Two curves are said to be touching if they have precisely one point in common, at which they do not properly cross.)
An important ingredient of our proofs is the following statement: Let $S$ be a family of the graphs of $n$ continuous real functions defined on $\mathbb{R}$, no three of which pass through the same point. If there are $nt$ pairs of touching curves in $S$, then the number of crossing points is $Ω(nt\sqrt{\log t/\log\log t})$.
△ Less
Submitted 20 December, 2014;
originally announced December 2014.
-
Kinetic Voronoi Diagrams and Delaunay Triangulations under Polygonal Distance Functions
Authors:
Pankaj K. Agarwal,
Haim Kaplan,
Natan Rubin,
Micha Sharir
Abstract:
Let $P$ be a set of $n$ points and $Q$ a convex $k$-gon in ${\mathbb R}^2$. We analyze in detail the topological (or discrete) changes in the structure of the Voronoi diagram and the Delaunay triangulation of $P$, under the convex distance function defined by $Q$, as the points of $P$ move along prespecified continuous trajectories. Assuming that each point of $P$ moves along an algebraic trajecto…
▽ More
Let $P$ be a set of $n$ points and $Q$ a convex $k$-gon in ${\mathbb R}^2$. We analyze in detail the topological (or discrete) changes in the structure of the Voronoi diagram and the Delaunay triangulation of $P$, under the convex distance function defined by $Q$, as the points of $P$ move along prespecified continuous trajectories. Assuming that each point of $P$ moves along an algebraic trajectory of bounded degree, we establish an upper bound of $O(k^4nλ_r(n))$ on the number of topological changes experienced by the diagrams throughout the motion; here $λ_r(n)$ is the maximum length of an $(n,r)$-Davenport-Schinzel sequence, and $r$ is a constant depending on the algebraic degree of the motion of the points. Finally, we describe an algorithm for efficiently maintaining the above structures, using the kinetic data structure (KDS) framework.
△ Less
Submitted 18 April, 2014;
originally announced April 2014.
-
On Kinetic Delaunay Triangulations: A Near Quadratic Bound for Unit Speed Motions
Authors:
Natan Rubin
Abstract:
Let $P$ be a collection of $n$ points in the plane, each moving along some straight line at unit speed. We obtain an almost tight upper bound of $O(n^{2+ε})$, for any $ε>0$, on the maximum number of discrete changes that the Delaunay triangulation $\mathbb{DT}(P)$ of $P$ experiences during this motion. Our analysis is cast in a purely topological setting, where we only assume that (i) any four poi…
▽ More
Let $P$ be a collection of $n$ points in the plane, each moving along some straight line at unit speed. We obtain an almost tight upper bound of $O(n^{2+ε})$, for any $ε>0$, on the maximum number of discrete changes that the Delaunay triangulation $\mathbb{DT}(P)$ of $P$ experiences during this motion. Our analysis is cast in a purely topological setting, where we only assume that (i) any four points can be co-circular at most three times, and (ii) no triple of points can be collinear more than twice; these assumptions hold for unit speed motions.
△ Less
Submitted 8 December, 2013;
originally announced December 2013.
-
On topological changes in the Delaunay triangulation of moving points
Authors:
Natan Rubin
Abstract:
Let $P$ be a collection of $n$ points moving along pseudo-algebraic trajectories in the plane. One of the hardest open problems in combinatorial and computational geometry is to obtain a nearly quadratic upper bound, or at least a subcubic bound, on the maximum number of discrete changes that the Delaunay triangulation $\DT(P)$ of $P$ experiences during the motion of the points of $P$.
In this p…
▽ More
Let $P$ be a collection of $n$ points moving along pseudo-algebraic trajectories in the plane. One of the hardest open problems in combinatorial and computational geometry is to obtain a nearly quadratic upper bound, or at least a subcubic bound, on the maximum number of discrete changes that the Delaunay triangulation $\DT(P)$ of $P$ experiences during the motion of the points of $P$.
In this paper we obtain an upper bound of $O(n^{2+\eps})$, for any $\eps>0$, under the assumptions that (i) any four points can be co-circular at most twice, and (ii) either no triple of points can be collinear more than twice, or no ordered triple of points can be collinear more than once.
△ Less
Submitted 12 April, 2013;
originally announced April 2013.
-
Statistical Topic Models for Multi-Label Document Classification
Authors:
Timothy N. Rubin,
America Chambers,
Padhraic Smyth,
Mark Steyvers
Abstract:
Machine learning approaches to multi-label document classification have to date largely relied on discriminative modeling techniques such as support vector machines. A drawback of these approaches is that performance rapidly drops off as the total number of labels and the number of labels per document increase. This problem is amplified when the label frequencies exhibit the type of highly skewed…
▽ More
Machine learning approaches to multi-label document classification have to date largely relied on discriminative modeling techniques such as support vector machines. A drawback of these approaches is that performance rapidly drops off as the total number of labels and the number of labels per document increase. This problem is amplified when the label frequencies exhibit the type of highly skewed distributions that are often observed in real-world datasets. In this paper we investigate a class of generative statistical topic models for multi-label documents that associate individual word tokens with different labels. We investigate the advantages of this approach relative to discriminative models, particularly with respect to classification problems involving large numbers of relatively rare labels. We compare the performance of generative and discriminative approaches on document labeling tasks ranging from datasets with several thousand labels to datasets with tens of labels. The experimental results indicate that probabilistic generative models can achieve competitive multi-label classification performance compared to discriminative methods, and have advantages for datasets with many labels and skewed label frequencies.
△ Less
Submitted 9 November, 2011; v1 submitted 13 July, 2011;
originally announced July 2011.
-
Kinetic Stable Delaunay Graphs
Authors:
Pankaj K. Agarwal,
Jie Gao,
Leonidas J. Guibas,
Haim Kaplan,
Vladlen Koltun,
Natan Rubin,
Micha Sharir
Abstract:
We consider the problem of maintaining the Euclidean Delaunay triangulation $\DT$ of a set $P$ of $n$ moving points in the plane, along algebraic trajectories of constant description complexity. Since the best known upper bound on the number of topological changes in the full $\DT$ is nearly cubic, we seek to maintain a suitable portion of it that is less volatile yet retains many useful propertie…
▽ More
We consider the problem of maintaining the Euclidean Delaunay triangulation $\DT$ of a set $P$ of $n$ moving points in the plane, along algebraic trajectories of constant description complexity. Since the best known upper bound on the number of topological changes in the full $\DT$ is nearly cubic, we seek to maintain a suitable portion of it that is less volatile yet retains many useful properties. We introduce the notion of a stable Delaunay graph, which is a dynamic subgraph of the Delaunay triangulation. The stable Delaunay graph (a) is easy to define, (b) experiences only a nearly quadratic number of discrete changes, (c) is robust under small changes of the norm, and (d) possesses certain useful properties.
The stable Delaunay graph ($\SDG$ in short) is defined in terms of a parameter $α>0$, and consists of Delaunay edges $pq$ for which the angles at which $p$ and $q$ see their Voronoi edge $e_{pq}$ are at least $α$. We show that (i) $\SDG$ always contains at least roughly one third of the Delaunay edges; (ii) it contains the $β$-skeleton of $P$, for $β=1+Ω(α^2)$; (iii) it is stable, in the sense that its edges survive for long periods of time, as long as the orientations of the segments connecting (nearby) points of $P$ do not change by much; and (iv) stable Delaunay edges remain stable (with an appropriate redefinition of stability) if we replace the Euclidean norm by any sufficiently close norm.
In particular, we can approximate the Euclidean norm by a polygonal norm (namely, a regular $k$-gon, with $k=Θ(1/α)$), and keep track of a Euclidean $\SDG$ by maintaining the full Delaunay triangulation of $P$ under the polygonal norm.
We describe two kinetic data structures for maintaining $\SDG$. Both structures use $O^*(n)$ storage and process $O^*(n^2)$ events during the motion, each in $O^*(1)$ time.
△ Less
Submitted 4 April, 2011;
originally announced April 2011.
-
Improved Bounds for Geometric Permutations
Authors:
Natan Rubin,
Haim Kaplan,
Micha Sharir
Abstract:
We show that the number of geometric permutations of an arbitrary collection of $n$ pairwise disjoint convex sets in $\mathbb{R}^d$, for $d\geq 3$, is $O(n^{2d-3}\log n)$, improving Wenger's 20 years old bound of $O(n^{2d-2})$.
We show that the number of geometric permutations of an arbitrary collection of $n$ pairwise disjoint convex sets in $\mathbb{R}^d$, for $d\geq 3$, is $O(n^{2d-3}\log n)$, improving Wenger's 20 years old bound of $O(n^{2d-2})$.
△ Less
Submitted 19 July, 2010;
originally announced July 2010.
-
A Kinetic Triangulation Scheme for Moving Points in The Plane
Authors:
Haim Kaplan,
Natan Rubin,
Micha Sharir
Abstract:
We present a simple randomized scheme for triangulating a set $P$ of $n$ points in the plane, and construct a kinetic data structure which maintains the triangulation as the points of $P$ move continuously along piecewise algebraic trajectories of constant description complexity. Our triangulation scheme experiences an expected number of $O(n^2β_{s+2}(n)\log^2n)$ discrete changes, and handles them…
▽ More
We present a simple randomized scheme for triangulating a set $P$ of $n$ points in the plane, and construct a kinetic data structure which maintains the triangulation as the points of $P$ move continuously along piecewise algebraic trajectories of constant description complexity. Our triangulation scheme experiences an expected number of $O(n^2β_{s+2}(n)\log^2n)$ discrete changes, and handles them in a manner that satisfies all the standard requirements from a kinetic data structure: compactness, efficiency, locality and responsiveness. Here $s$ is the maximum number of times where any specific triple of points of $P$ can become collinear, $β_{s+2}(q)=λ_{s+2}(q)/q$, and $λ_{s+2}(q)$ is the maximum length of Davenport-Schinzel sequences of order $s+2$ on $n$ symbols. Thus, compared to the previous solution of Agarwal et al.~\cite{AWY}, we achieve a (slightly) improved bound on the number of discrete changes in the triangulation. In addition, we believe that our scheme is simpler to implement and analyze.
△ Less
Submitted 6 May, 2010;
originally announced May 2010.
-
Line Transversals of Convex Polyhedra in $\reals^3$
Authors:
Haim Kaplan,
Natan Rubin,
Micha Sharir
Abstract:
We establish a bound of $O(n^2k^{1+\eps})$, for any $\eps>0$, on the combinatorial complexity of the set $\T$ of line transversals of a collection $¶$ of $k$ convex polyhedra in $\reals^3$ with a total of $n$ facets, and present a randomized algorithm which computes the boundary of $\T$ in comparable expected time. Thus, when $k\ll n$, the new bounds on the complexity (and construction cost) of…
▽ More
We establish a bound of $O(n^2k^{1+\eps})$, for any $\eps>0$, on the combinatorial complexity of the set $\T$ of line transversals of a collection $¶$ of $k$ convex polyhedra in $\reals^3$ with a total of $n$ facets, and present a randomized algorithm which computes the boundary of $\T$ in comparable expected time. Thus, when $k\ll n$, the new bounds on the complexity (and construction cost) of $\T$ improve upon the previously best known bounds, which are nearly cubic in $n$.
To obtain the above result, we study the set $\TL$ of line transversals which emanate from a fixed line $\ell_0$, establish an almost tight bound of $O(nk^{1+\eps})$ on the complexity of $\TL$, and provide a randomized algorithm which computes $\TL$ in comparable expected time. Slightly improved combinatorial bounds for the complexity of $\TL$, and comparable improvements in the cost of constructing this set, are established for two special cases, both assuming that the polyhedra of $¶$ are pairwise disjoint: the case where $\ell_0$ is disjoint from the polyhedra of $¶$, and the case where the polyhedra of $¶$ are unbounded in a direction parallel to $\ell_0$.
△ Less
Submitted 8 July, 2008;
originally announced July 2008.
